# Getting Started
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
MeshMesh is currently invite only. If you'd like to try it out, [join our early access program](https://meshmesh.io/?early-access=true).
## Download and Install
Once you've received an invite via email, download MeshMesh from [console.meshmesh.io/download](https://console.meshmesh.io/download). We currently support MacOS — Windows support is coming soon.
## First Steps
After installing, sign in with your MeshMesh account. You'll land on the dashboard where you can create your first agent.
MeshMesh includes a demo Salesforce org so you can explore right away without connecting your own environment. When you're ready to work with your own data, connect your Salesforce org from the Integrations menu.
# Overview
import { Card, Cards } from 'fumadocs-ui/components/card';
MeshMesh Studio is a new way to be extraordinarily productive. Build AI agents, analyze your business, create campaigns and content — in minutes instead of weeks.
With deep integrations across Salesforce and other platforms, MeshMesh handles the complexity so you can focus on getting things done.
## Why MeshMesh
MeshMesh is custom-built for Salesforce and adjacent products. Unlike general-purpose AI tools, our context model understands Salesforce data structures, APIs, and workflows deeply — delivering the right value for your specific environment.
## What You Can Do
Here are some examples of what you can accomplish. See more on the [use cases](/use-cases) page.
Create and deploy AI agents within your Salesforce environment. **85-90% time saved**
Get insights on pipeline, performance, and opportunities across your data. **80-85% time saved**
Generate emails, landing pages, and marketing assets on-brand. **85-90% time saved**
Configure leads, accounts, opportunities, and sales processes. **80-90% time saved**
Design and deploy multi-channel customer journeys. **85-90% time saved**
Unify customer data and build targeted segments. **80-85% time saved**
## Supported Products
Coverage varies by product. We're continuously expanding support.
### Marketing
* Marketing Cloud Engagement
* Account Engagement (Pardot)
* Loyalty Management
### Commerce
* B2C Commerce (experimental)
* B2B Commerce & D2C Commerce (experimental)
* Composable Storefront (experimental)
* CPQ (experimental)
* Subscription Management (experimental)
### Sales & Service
* Sales Cloud
* Service Cloud
* Field Service
* Messaging & Conversation Service
### Data & Analytics
* Data Cloud
* CRM Analytics (experimental)
* Tableau (experimental)
* Reports & Dashboards (experimental)
### AI & Automation
* Agentforce
* Einstein Bots (experimental)
* Einstein Discovery (experimental)
* Einstein Vision & Language (experimental)
* Models API (LLMs) (experimental)
* Flows (experimental)
### Industry Clouds
* Financial Services Cloud (experimental)
* Healthcare (experimental)
* Manufacturing Cloud (experimental)
* Consumer Goods Cloud (experimental)
* Energy & Utilities Cloud (experimental)
* Communications Cloud (experimental)
* Public Sector (experimental)
* Net Zero Cloud (experimental)
See [use cases](https://meshmesh.io/usecases) for details and more. Don't see your product? Get in touch at [hello@meshmesh.io](mailto:hello@meshmesh.io).
# Network Allowlist
import { Callout } from 'fumadocs-ui/components/callout';
When MeshMesh agents run tasks, they make outbound API calls to Salesforce, Marketing Cloud, your custom endpoints, and other services on your behalf. All of this traffic originates from a small, fixed set of static IP addresses, so your security team can allowlist them in firewalls, API gateways, and IP-restricted Salesforce orgs.
## IP addresses to allowlist
| IP Address | Region |
| ---------------- | ------- |
| `209.71.108.40` | US East |
| `209.71.102.179` | US East |
**Allowlist all listed IPs.** MeshMesh traffic to your systems may originate from any of them. Allowlisting only some can cause intermittent failures.
## Recommended allowlist entries
Use individual `/32` entries — these IPs are not contiguous:
```
209.71.108.40/32
209.71.102.179/32
```
## Stability guarantees
* **IPs are stable.** Reserved at the cloud level and persist across host failures and infrastructure rotations. We do not change them without prior notice.
* **We will notify you** at least 30 days in advance via email if any IP needs to change for any reason.
* **We may add IPs** as we scale (additional regions, increased capacity). New IPs will be announced ahead of time. Existing IPs will continue to work.
## Verifying connectivity
To confirm your allowlist is working, run a test task that calls your endpoint and check your edge logs for inbound traffic from one of the listed IPs. If you'd like help verifying the setup, contact your MeshMesh representative.
## Regions
MeshMesh currently runs in **US East**. If your security policy requires data to remain in a specific region (EU, APAC, etc.), reach out — we can deploy dedicated regional infrastructure.
## Frequently asked questions
**Are these IPs shared with other MeshMesh customers?**
Yes — they are shared infrastructure. Your traffic is authenticated end-to-end (TLS to your endpoints) and never co-mingled with other tenants' data. The shared IPs are only the egress address.
**What if my organization's security policy doesn't allow shared egress IPs?**
Contact your MeshMesh representative — we can discuss dedicated egress IPs for enterprise customers.
# Release Notes
RSS Feed
## August 26, 2026
* **Reach systems behind enterprise IP allowlists** — agents now connect through stable, dedicated outbound IP addresses, so they can work with **GitHub Enterprise**, **Marketing Cloud**, and **IP-restricted Salesforce orgs** that only accept traffic from approved addresses. Your IT team allowlists MeshMesh's addresses once and agents can securely reach those systems
[Read more →](/releases/2026-08-26)
## August 25, 2026
* **Faster, more reliable Analytics** — the Analytics → Users view now loads from precomputed usage data instead of calculating everything on the fly, so it opens quickly and reliably even on large workspaces (it previously could time out and show "No users yet"). If it can't load, you now get a clear error with a **Retry** button instead of an empty state
* **More member details, and a roster you can pull with agents** — the Users view now shows each member's role and join date, and a new admin capability lets agents retrieve your full workspace member list, including people who haven't been active yet
[Read more →](/releases/2026-08-25)
## August 24, 2026
* **Connect your favorite tools to agents over MCP** — you can now connect remote MCP servers with a secure sign-in (OAuth), including **Linear, Notion, Atlassian, Tableau, and Heroku** — or any custom MCP server by pasting its URL — so agents can use their tools and data right inside a task. Connect once, give each connection a name, and turn it on or off anytime
* **More reliable Salesforce connections** — when a Salesforce connection can no longer refresh its access, MeshMesh now clearly asks you to reconnect and explains why, instead of quietly failing partway through a task
* **Security & reliability improvements** — some behind-the-scenes security hardening
[Read more →](/releases/2026-08-24)
## August 4, 2026
* **Fixed a sign-in loop that could affect newly invited members** — in some cases an invited teammate could be signed out shortly after signing in, until an admin re-enabled their access; invited members now stay signed in and reach their workspace the first time
* **Better at keeping you in the workspace you selected** — in some cases switching workspaces on desktop could snap you back to the one you signed in with; your choice now stays put more reliably, even when the app refreshes your sign-in in the background
* **Security & reliability improvements** — dependency security patches and behind-the-scenes hardening
[Read more →](/releases/2026-08-04)
## July 28, 2026
* **Fixed an issue where output downloads could fail** — downloads no longer depend on the task's live workspace, so they no longer fail with "Failed to download file" when the task has gone idle; outputs render from your saved content instead
* **Fewer dropped messages during very busy tasks** — fixed a database issue that could occasionally fail to record a message under heavy load
* **Security & reliability improvements** — dependency security patches and behind-the-scenes hardening
[Read more →](/releases/2026-07-28)
## July 24, 2026
* **Connect Salesforce with your own External Client App** — a new connection method lets you connect Salesforce Core using an External Client App you create, so agents run on the web and keep working on their own — including scheduled and always-on tasks — with no desktop app needed. Especially useful on Windows and the web version
[Read more →](/releases/2026-07-24)
## July 22, 2026
* **A clearer Generated Outputs list** — a sortable, newest-first list showing when each output was created and which agent made it; sort by newest / name / type and filter by type, persisted per workbench
* **Consistent actions on every item** — rename, star, download, remove, or save to your library from hover or right-click, the same way in list and grid
* **Save an output to your library** — promote a great output into a reusable reference in a couple of clicks
* **Folders for references and outputs** — group a task's items into folders: create them, drag items in, add or remove via right-click, and open a folder to focus on its contents
* **Download on the web, and in bulk** — the output viewer now downloads on the web too, and you can bulk-download several items as one ZIP
* **Find within a document (⌘F)** — search inside plans, references, and outputs, jumping between matches with a match count
* **A cleaner references picker** — descriptions and usage counts on every row, persistent filters, a fixed search-with-filter trap, and description/type/tags captured when you create a reference
* **Security & reliability improvements** — high-severity dependency security patches and behind-the-scenes hardening
[Read more →](/releases/2026-07-22)
## June 16, 2026
* **Faster startup and load times** — improved startup performance on the desktop app, so it gets you to your tasks and workspace quicker on refresh or relaunch, with the latest data filling in as it loads
* **Remove a task's outputs** — you can now delete outputs (artifacts) from a task, just like you can with references, to keep a task tidy
* **Fixed a rare screen freeze on confirmations** — when an agent asked you to confirm a sensitive action, another panel could occasionally open on top and lock the screen; the confirmation now always stays reachable, with a clear way to dismiss it
* **Security & reliability improvements** — high-severity dependency security patches and behind-the-scenes hardening
[Read more →](/releases/2026-06-16)
## June 12, 2026
* **Live-updating file browser** — the agent's file browser now refreshes the moment files change, using a real-time signal instead of polling every few seconds, so new and edited files show up without the wait
* **Fewer timeouts and failures during busy periods** — a series of behind-the-scenes database improvements keep tasks starting, resuming, and browsing files reliably when the system is under heavy load
* **Marketing Cloud connections stop spuriously asking to reconnect** — fixed a reauthentication loop that could repeatedly prompt some Marketing Cloud connections to reconnect; they now refresh quietly (the same fix we shipped for Salesforce on June 7)
* **Security & reliability improvements** — behind-the-scenes security hardening and dependency security patches
[Read more →](/releases/2026-06-12)
## June 7, 2026
* **More reliable Salesforce connections — with a clear way back when one needs attention** — fixed a problem where a failed Salesforce connection could leave an agent stuck "searching for the connection." Connections that need a refresh now show an amber reconnect badge in your Integrations list and the task picker, and a failed task shows a friendly banner with a one-click **Reconnect** button instead of stalling
* **Accurate Activity numbers** — the Activity dashboard's total-users count now reflects everyone on your team (it previously capped out and under-counted on larger workspaces), and the leaderboard updates reliably again
* **Security & reliability improvements** — behind-the-scenes security hardening and dependency security patches
[Read more →](/releases/2026-06-07)
## June 2, 2026
* **Business Intelligence opens cleanly** — opening Business Intelligence now shows a brief loading state and then your existing analyses, instead of flashing "No analyses yet" before they load
* **Correct inviter on team invitations** — invitations now record and show the right person as the inviter
* **Hardened desktop sign-in** — the desktop sign-in handoff is more secure behind the scenes; nothing changes about how you sign in
* **Security patches** — dependency updates addressing known advisories
[Read more →](/releases/2026-06-02)
## June 1, 2026
* **Richer Activity Reports** — task cards now show when they were last updated and their star rating, the leaderboard adds an Agent Time column, and Active Teammates counts everyone who has run an agent; cleaner, more consistent cards across the feed and profiles
* **Smarter, more up-to-date agents** — agents now check the latest product docs before reaching for a skill, so they apply the freshest guidance and give more accurate answers
[Read more →](/releases/2026-06-01)
## May 30, 2026
* **Task ID in Salesforce webhooks** — Salesforce webhook payloads now include the MeshMesh task ID, so a triggered task is easy to tie back to its source
* **Behind-the-scenes reliability** — quieter handling of transient sandbox hiccups, improved Data Cloud guidance for agents after a recent Salesforce CLI change, and a large expansion of automated test coverage; no change to day-to-day use
[Read more →](/releases/2026-05-30)
## May 29, 2026
* **Smoother member management in the Admin Console** — invites now work reliably with clear messages when someone's already invited or already a member; search your member list, re-invite pending invitations, and manage several members at once with new bulk actions
* **Fixed Salesforce agents getting stuck asking to reconnect** — resolved an issue affecting both Quick Connect and Custom App connections where agents could repeatedly ask you to reconnect and report that no orgs were found, even on a healthy connection; agents now pick up your Salesforce connection and get to work
* **Smoother Marketing Cloud reconnection on the web** — the sign-in window now opens and closes itself cleanly, and no longer reports success before you've finished signing in
* **Security & reliability improvements** — behind-the-scenes hardening, dependency security patches, a reduced log surface, and a large expansion of automated test coverage; no change to how you work day to day
[Read more →](/releases/2026-05-29)
## May 27, 2026
* **Review your agent's work without losing your place** — the task-complete review card minimizes into a small "Review Agent" pill, so it no longer covers your references and artifacts
* **Steadier sign-in on desktop** — hardened a few rare edge cases where the desktop app could drop your session; it now refreshes quietly and retries before signing out
* **Better Agentforce builder detection** — fixed some cases where agents would reach for the legacy builder; they now more reliably use the new Agentforce Builder (Agent Script) when that's what you want
[Read more →](/releases/2026-05-27)
## May 26, 2026
* **Monthly usage limits with a clear heads-up** — set a monthly workspace limit and get a gentle in-app warning as you approach it, instead of a task quietly failing
* **See your usage in the app** — a new usage view shows your workspace's monthly usage at a glance
* **Admin Console for workspace owners** — manage members, usage limits, allowed integrations, and an audit log in one place; deactivated members now appear in the members list
[Read more →](/releases/2026-05-26)
## May 23, 2026
* **Cleaner error messages** — raw system or usage-policy error text no longer leaks into the chat; you see friendly, readable copy instead
## May 19, 2026
* **Fixed Marketing Cloud "please reconnect" loop on long tasks** — agents 20+ min into a Marketing Cloud run no longer spuriously ask you to refresh the connection
* **Correct connection picker when a Marketing Cloud refresh is genuinely needed** — shows the right connection type and only matching connections
* **Auto-recovery from one-off Marketing Cloud sign-in errors** — agents self-heal from transient token hiccups without interrupting the task
* **Fixed typing special characters into the agent's browser** — `@`, `!`, `#`, `$` and other shifted symbols sometimes got dropped on login forms
* **Fixed browser thumbnail overlay overlapping workbench tabs** — the "Agent browsing" indicator now lives in the page header with a working live preview
[Read more →](/releases/2026-05-19)
## May 18, 2026
* **Reliability & security improvements** — internal hardening release: dependency security patches across the stack and a substantial expansion of automated test coverage on the API gateway and supporting services. No user-facing changes.
## May 15, 2026
* **Cloud Control 5.1 (Preview) — our largest context model yet**, now available in preview. Substantially improved coverage across Health Cloud, Education Cloud, Public Sector, Nonprofit, Consumer Goods, OmniStudio, Revenue Cloud, CRM Analytics, Tableau Next, Marketing Intelligence, Agentforce, Data Cloud, Platform, Field Service, Commerce Cloud, Experience Cloud, Marketing Cloud Engagement, and Service Cloud. New agents default to CC5.1.
* **Connect Salesforce sandboxes and custom domains via MeshMesh App** — fixes a bug where the **MeshMesh App** connection method would fail with "Grant type not supported" for sandbox or custom My Domain orgs.
[Read more →](/releases/2026-05-15)
## May 13, 2026
* **More reliable Salesforce connections** — fixes a bug where Salesforce integrations could repeatedly fail with "Connection credentials could not be loaded. Please reconnect.", especially when an org had multiple connections sharing the same Connected App. Affected connections will need one final reconnect, after which they should stay healthy.
[Read more →](/releases/2026-05-13)
## May 12, 2026
* **Faster agent browsing** — every click, snapshot, and page change goes through a streamlined pipeline, so long browser tasks feel noticeably snappier turn by turn
* **Up to 75% fewer tokens per browser turn** — the same browser work now uses far less context, so browser-using agents stay leaner on long tasks
[Read more →](/releases/2026-05-12)
## May 8, 2026
* **Improved page rendering in the agent's live browser** — Salesforce, Trailhead, Marketing Cloud, and other sign-in and signup pages now load cleanly in the live browser view
* **Correct aspect ratio in the live browser** — pages render at full 1280×800 instead of looking slightly horizontally squished
[Read more →](/releases/2026-05-08)
## May 7, 2026
* **Workbenches** — a new shared workspace where multiple agents collaborate on the same project, with shared connections, references, and direct hand-offs between agents
* **Plan mode by default** for new agents in a workbench, so they map out the work first
* **Watch each agent live** — every agent in a workbench has its own browser view
* Calmer create-workbench experience while your shared workspace is prepared
[Read more →](/releases/2026-05-07)
## May 6, 2026
* **Improved connection handling on the web** — Salesforce connection sign-in in the browser is more reliable end-to-end: opens in a popup, surfaces errors inline, closes itself on success, and no longer leaves half-finished connections behind on cancel
* **Tutorial videos now play on the web app** — onboarding, dashboard, and settings videos no longer get blocked by the browser
[Read more →](/releases/2026-05-06)
## May 5, 2026
* **MeshMesh Studio on the web** — sign in at [studio.meshmesh.io](https://studio.meshmesh.io) and use the full Studio in your browser
* **Smoother onboarding setup** — first-run "Connect Your Tools" step now uses the real Integrations flow, with proper Salesforce, Marketing Cloud, Slack, GitHub, and Mulesoft OAuth and server-side credential storage
* **Friendlier credits experience** — no more false "out of credits" warnings, a gentle low-credits heads-up, and past errors clear themselves once you top up
* **Marketing Cloud setup guide updated** — both desktop and web redirect URIs are now documented; admins running their own Installed Package can register both so it works in either app
* Cleaner integrations popover in chat, polished connection screens, setup-guide link on the Salesforce MeshMesh App connection method
[Read more →](/releases/2026-05-05)
## May 4, 2026
* **Turn integrations on or off** — toggle any integration in the Your Integrations modal to hide it from the agent prompt picker without deleting it
* **Delete legacy Marketing Cloud connections in one click** — the deprecation prompt now offers Delete and Delete & create new directly, no more digging through menus
[Read more →](/releases/2026-05-04)
## May 1, 2026
* **Marketing Cloud reconnection now Just Works** — expired or revoked sign-ins prompt you to reconnect up front, with details pre-filled, before the task runs
* **No duplicate connections on reconnect** — reconnecting updates the existing connection instead of creating a second copy
* **Long Marketing Cloud agents stay signed in** — background sign-in refreshes during long tasks no longer knock the agent offline
[Read more →](/releases/2026-05-01)
## April 28, 2026
* **Celebration no longer blocks the chat** — task completion card now overlays the left panel so the chat stays visible
* **More reliable tag editing** — tag suggestions add on first click; sidebar tag filter is a clean single-select
* **More reliable brand kit creation** — steadier web scraping for brand kit setup
* Stability and reliability improvements
[Read more →](/releases/2026-04-28)
## April 27, 2026
* **New Public App integration for Marketing Cloud** — Marketing Cloud now connects via a Public App with personal sign-in, replacing the previous shared-credentials setup. Existing Marketing Cloud connections will be deactivated and need to be set up again following our [new connection guide](/guides/connect-marketing-cloud).
[Read more →](/releases/2026-04-27)
## April 21, 2026
* **Cloud Control 5 (Preview)** — opt into our latest, largest Cloud Control for noticeably better coverage across more of the Salesforce platform
* **Pick per agent** — choose your Cloud Control version when starting a new agent
* **Always know what's powering your agent** — every agent now shows a small badge indicating the Cloud Control version it's using
* **Smoother file drops in file view** — drop files directly onto a folder, file, or empty space in the file browser to upload exactly where you want, without the reference uploader getting in the way
* **Uninterrupted B2C Commerce Cloud sessions** — long-running Commerce Cloud agents no longer stall around the 30-minute mark; authentication now refreshes quietly in the background so tasks keep going
[Read more →](/releases/2026-04-21)
## April 15, 2026
* **Business Intelligence** — agent-powered org analysis with rich dashboards, health scoring, charts, heatmaps, and actionable recommendations
* **Enhanced file browser** — integrated file browsing alongside the chat, drag-and-drop uploads with ZIP support, file management, and live agent edit tracking
* **Drag-and-drop on dashboard** — drop images and files directly into the prompt area when starting a new task
* **Improved B2C Commerce Cloud connections** — clearer credential errors and automatic token refresh
* Security and reliability improvements
[Read more →](/releases/2026-04-15)
## April 8, 2026
* **Slack Enterprise Grid support** — connect Slack workspaces within Enterprise Grid organizations, with automatic workspace routing
* **Improved reference editing** — search and find-replace in code and markdown editors, plus a unified editing experience
* Stability and performance improvements
[Read more →](/releases/2026-04-08)
## April 3, 2026
* **Fewer interruptions** — agent sessions are more stable, resolving occasional "Failed to send" errors
* **Additional guardrails around browser use** — extra checks when agents browse to keep sessions running smoothly
* Faster response times, smoother reconnections
[Read more →](/releases/2026-04-03)
## April 1, 2026
* **Better connection error handling** — integrations now detect and resolve common authentication and network issues automatically, including corporate proxy and VPN setups
* **More resilient agents** — agents now stop after repeated denied actions instead of looping, and lost connections pause the task so you can resume
* Stability and performance improvements
[Read more →](/releases/2026-04-01)
## March 31, 2026
* **Reference list view & search** — browse references in a compact list and instantly find what you need by searching names, descriptions, or content
* **Editable time saved** — adjust your agent's time-saved estimate and share feedback after every task
* Stability and reliability improvements
[Read more →](/releases/2026-03-31)
## March 29, 2026
* **Slack integration** — bring your own Slack apps and bots, and let MeshMesh orchestrate them — send messages, manage channels, and run workflows without leaving the app
* Longer sessions, connection reliability improvements
[Read more →](/releases/2026-03-29)
## March 27, 2026
* **Smart action escalation** — your agent now automatically detects critical actions and requests your confirmation before proceeding, giving you an extra safety harness and peace of mind
* **B2C Commerce Cloud improvements** — more reliable and secure B2C Commerce connections
* **Longer connection names** — Salesforce connection names now support more characters
* Reliability and security improvements
[Read more →](/releases/2026-03-27)
## March 26, 2026
* **B2C Commerce Cloud (Beta)** — connect your Commerce Cloud instance for code deployment, shopper login, on-demand sandboxes, and storefront management
* **Auto storage resize** — agent storage upgrades automatically when disk fills up
* **Live progress** — see real-time elapsed time and line counts during agent work
* **Faster streaming** — long agent sessions no longer timeout
* Better failover, more reliable sign-up, search docs
[Read more →](/releases/2026-03-26)
## March 24, 2026
* **Archive agents** — hide agents you're not using from the sidebar without deleting them
* **Success celebrations** — get visual confirmation when your agent completes work, with highlights of what was done
* **What's new on launch** — see what you missed since your last visit with an in-app changelog
* **Browser activity indicator** — always know when your agent is browsing, even with overlays open
* **Quick-add references** — recent and starred references appear as pills for instant one-click adding
* Clearer integrations layout, fewer unnecessary credential prompts, skills enabled by default
[Read more →](/releases/2026-03-24)
## March 23, 2026
* **Recipe creator** — turn any completed agent run into a reusable recipe your team can follow
* **Reference sharing & organization** — share references with your team using Google Drive-style permissions, star your favorites, and filter by type or tag
* **@ mention references** — type @ in chat to quickly pull in any reference as context for your agent
* **Agent image sharing** — your agent can now upload and use images during tasks, with inline previews and download
* **Brand kit previews** — see live visual previews of brand kit content with auto-cycling examples
* Improved search, faster reference browsing, fullscreen artifact preview
[Read more →](/releases/2026-03-23)
## March 16, 2026
* **Agent rename** — rename agents directly from the sidebar context menu
* **More file types** — upload JSON, YAML, HTML, and audio files as references
* **Auto-recovery** — agent environments automatically restore from backup if interrupted
* Faster search, Salesforce setup guide, security patches
[Read more →](/releases/2026-03-16)
## March 13, 2026
* **More stable long-running tasks** — agents handle extended conversations without crashing
* Improved agent reliability, more reliable streaming
* Fixed file downloads for agent-generated assets
[Read more →](/releases/2026-03-13)
## March 12, 2026
* **More reliable Salesforce connections** — connections properly await provisioning, preventing timeouts
* Improved sandbox stability, security hardening
[Read more →](/releases/2026-03-12)
## March 11, 2026
* **Prompt navigation** — Cmd/Ctrl+Up/Down to jump between messages
* **Task list pagination** — browse older tasks with "Load more"
* Demo connection protection, sandbox stability, search fixes
[Read more →](/releases/2026-03-11)
## March 10, 2026
* **Cross-agent keyword search** — search across all your agents and messages with Cmd+K
* More accurate usage tracking
[Read more →](/releases/2026-03-10)
## March 5, 2026
* **Inline screenshot sharing** — share screenshots directly in task chat
* **Smarter agent evaluation** — agents evaluate their own work with structured checkpoints
* File preview fixes, PDF export fixes, UI improvements
[Read more →](/releases/2026-03-05)
## March 4, 2026
* **Selectable artifact text** — select and copy text from artifacts and references
* **Long session recall** — agents reference earlier artifacts in extended sessions
* Smarter chat scrolling, individual demo environments
* Security improvements
[Read more →](/releases/2026-03-04)
## February 24, 2026
* **Built-in Slack support** — get help directly from the app
* Fixed chat input carrying over between tasks
* Security improvements
[Read more →](/releases/2026-02-24)
## February 23, 2026
* Better handling of large PDF uploads
* Performance and reliability improvements
[Read more →](/releases/2026-02-23)
## February 20, 2026
* HTML artifacts now download as `.html` files for better compatibility
* Fixed artifact display issues, improved agent stability
[Read more →](/releases/2026-02-20)
## Version 0.20.9
**February 19, 2026**
Your agent can now use a real browser — navigating Salesforce, clicking through Setup, and interacting with pages while you watch live.
* **Browser Use** - agents navigate and interact with Salesforce UI in a live browser view
* **GitHub Integration** - connect private repos for agent access
* **Video Tutorials** - contextual tutorial videos embedded throughout the app
* Restyled login, smarter plan mode, improved agent reliability
* Fixed Salesforce credential timeouts and chat stability issues
[Read more →](/releases/0-20-9)
## Version 0.20.4
**February 3, 2026**
* Salesforce sandbox support - connect to sandbox environments for testing
* Bulk task management - select and delete multiple tasks at once
* Smoother agent conversations and improved connection reliability
[Read more →](/releases/0-20-4)
## Version 0.20.3
**January 28, 2026**
* Fixed login issues on first launch
* Fixed team invitation flow
## Version 0.20.0
**January 27, 2026**
Our biggest release yet - introducing Blueprints for end-to-end implementations, a completely revamped interface, new skills for Salesforce clouds, and demo connections to get started instantly.
* **Blueprints** - End-to-end implementation guides for Data 360, Agentforce agents, Marketing Cloud campaigns, and more
* **Revamped UX** - Cleaner dashboard with task list, context cards, and improved task experience
* **New Agent Skills** - Agentforce, Marketing Cloud, Data Cloud, and metadata deployment
* **BYOK** - Bring your own API keys with enterprise controls
* **Demo Connections** - Try Salesforce and Marketing Cloud instantly
* **Public References** - Share and discover references across your organization
* **Claude Opus 4.5** - Support for the latest Claude model
[Read more →](/releases/0-20-0)
## Version 0.19.31
**January 10, 2026**
* Simplified onboarding - all connections auto-selected by default
* Salesforce connection now optional during setup
* Fixed an issue with connecting to Salesforce on Mac
[Read more →](/releases/0-19-31)
## Version 0.19.30
**January 8, 2026**
* Intel Mac support - Desktop app now available for Intel-based Macs
* Auto-update on quit - App updates automatically in the background
* Improved workspace switching
* Fixed plan mode switching issues
[Read more →](/releases/0-19-30)
## Version 0.19.28
**January 7, 2026**
* Better reference management - improved selection and direct editing
* Fixed light mode styling
* Fixed reference upload issues
[Read more →](/releases/0-19-28)
## Version 0.19.22
**December 25, 2025**
* Interactive questions - AI can ask for clarification during tasks
* Plan mode enabled by default
* Fixed artifact display issues in plan mode
* Fixed issue with updating plans
[Read more →](/releases/0-19-22)
## Version 0.19.21
**December 12, 2025**
* Plan mode improvements - better visual mode switching and smarter planning
* PDF and CSV file support - upload documents directly to agents
* Fixed message streaming issues when tasks stop
* Improved offline banner display
* Security updates
[Read more →](/releases/0-19-21)
## Version 0.19.20
**December 11, 2025**
* Improved login button styling
## Version 0.19.19
**December 11, 2025**
* Agent coding abilities - AI can now write and execute code during tasks
* Salesforce custom scopes - configure CDP and Ingest API permissions
[Read more →](/releases/0-19-19)
## Version 0.19.18
**December 11, 2025**
* Fixed file upload issues in agent sandbox
[Read more →](/releases/0-19-18)
## Version 0.19.17
**December 11, 2025**
* Planning mode - AI now plans before making changes for more accurate results
* Beta release channel for early access to new features
[Read more →](/releases/0-19-17)
## Version 0.19.13
**December 10, 2025**
* Live task context - change connections and references while a task is running
* Faster reference loading with smart caching
* Fixed Salesforce connection issues after periods of inactivity
[Read more →](/releases/0-19-13)
## Version 0.19.11
**December 9, 2025**
* Brand kit creation during onboarding - set up your brand in the first run experience
* Workspace naming during setup
* Improved brand kit preview
* Security improvements
[Read more →](/releases/0-19-11)
## Version 0.19.10
**December 6, 2025**
* Fixed agent connection issues when restarting
[Read more →](/releases/0-19-10)
## Version 0.19.8
**December 5, 2025**
* File browser - browse and view files created by the AI agent during tasks
* Fixed agent not restarting correctly
* Fixed custom MCP server import issues
[Read more →](/releases/0-19-8)
## Version 0.19.7
**December 4, 2025**
* Task subagents - AI can spawn sub-tasks for deeper exploration and research
* Fixed actions not displaying after page refresh
[Read more →](/releases/0-19-7)
## Version 0.19.6
**December 4, 2025**
* Reference sharing - share references with team members
[Read more →](/releases/0-19-6)
## Version 0.19.5
**December 4, 2025**
* Fixed Marketing Cloud MCP integration issues
* Fixed stop button now shows progress immediately
## Version 0.19.3
**December 3, 2025**
* AI-created references - agent can automatically create and update reference documents during tasks
* Custom business context - pass brand kits and context to content creation
* Fixed HTML artifacts not displaying correctly on page reload
## Version 0.19.2
**December 2, 2025**
* Team invites - invite team members to collaborate in your workspace
## Version 0.19.0
**December 1, 2025**
* Workspaces - organize work by team or project with multi-tenant support
* Credits are now tracked per workspace instead of per user
## Version 0.18.34
**November 26, 2025**
* Brand kit import - import brand kits from external sources
* Fixed MCP connection credentials not being passed correctly to servers
* Fixed image and file display for authenticated API endpoints
* Fixed file upload authentication when dropping files into chat
## Version 0.18.1
**November 20, 2025**
* Fixed Marketing Cloud SOAP API token refresh - connections now automatically refresh expired tokens
## Version 0.18.0
**November 17, 2025**
* Made Reference Management more intuitive
* Introducing Auto-Business Context for some early alpha users
## Version 0.17.0
**November 5, 2025**
* Reference file tree management - organize documents with folders and nested files
* Full file operations - create, rename, delete, and move files
## Version 0.16.0
**November 3, 2025**
* Early tool call feedback - see tool activity instantly without waiting
## Version 0.15.0
**October 28, 2025**
* Starred artifacts dashboard - save and view your favorite artifacts on the dashboard
## Version 0.14.0
**October 23, 2025**
* Brand kit templates - apply branded templates to AI-generated content
* Persona feedback - get feedback from different persona perspectives
## Version 0.13.0
**October 20, 2025**
* Auto-documentation - AI automatically documents learnings and best practices
* File drop support - drag and drop files directly into tasks
## Version 0.12.0
**October 17, 2025**
* MuleSoft integration - connect to MuleSoft Anypoint Platform
* Flexible MCP configuration - easily add any MCP server with JSON config
* Asset display - AI can show images, videos, PDFs, and other media files
## Version 0.11.0
**October 13, 2025**
* HTML artifacts - AI can create rich interactive HTML content
* Safe HTML rendering with security protections
## Version 0.10.0
**October 12, 2025**
* Salesforce Marketing Cloud integration - connect to SFMC with AI assistance
## Version 0.9.0
**October 11, 2025**
* Task templates - one-click task creation with pre-configured templates
* Improved navigation with global sidebar for tasks and references
* Document editor with markdown support
## Version 0.8.0
**October 9, 2025**
* Enhanced Slack bot with rich formatting and artifact display
## Version 0.7.0
**October 8, 2025**
* Artifacts - AI automatically documents insights, reports, and challenges during work
* View artifacts by category: insights, reports, challenges, questions, decisions
* Visual displays with charts, tables, and callouts
* Real-time artifact updates as AI discovers new information
## Version 0.6.0
**October 7, 2025**
* Faster action feedback - see what's happening in real-time with loading states
* Redesigned task view with cleaner layout and better action visibility
* Smart TODO display that adapts based on task progress
## Version 0.5.0
**October 6, 2025**
* Action intelligence - see what AI is doing in plain language, not technical commands
## Version 0.4.0
**October 2, 2025**
* Performance improvements - smoother experience with large chat histories
## Version 0.3.0
**September 22, 2025**
* Cloud Control v2 - intelligent documentation reconciliation system
* Automated documentation synchronization
* Query your documentation directly
## Version 0.2.0
**September 5, 2025**
* Task ID display - easily identify tasks in the UI
## Version 0.1.0
**September 2, 2025**
* Desktop app for Cloud Control Studio - run everything locally on your machine
* Configurable workspace to organize all your work in one place
* Your data stays on your computer with automatic file-based storage
## Version 0.0.2
**September 1, 2025**
* Slack integration - create and manage tasks directly from Slack
* `/intent` command to create new tasks with interactive modal
* `/intent-refs` command to list your available references
* `/intent-status` command to check task progress
## Version 0.0.1
**August 30, 2025**
* Initial release of Cloud Control Studio
* Core task management system
* Reference management
# Security
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
import { Tabs, Tab } from 'fumadocs-ui/components/tabs';
import { Accordion, Accordions } from 'fumadocs-ui/components/accordion';
import { Steps } from 'fumadocs-ui/components/steps';
MeshMesh is built with security and privacy as foundational principles. Your Salesforce data never leaves the secure connection between MeshMesh and Salesforce.
We use bank-level encryption and industry-leading security practices across all Salesforce products including Marketing Cloud Engagement, Sales Cloud, Service Cloud, and Data Cloud.
## Security Architecture
|TLS 1.3| B[MeshMesh]
B -->|AES-256| C[Encrypted Storage]
B -->|OAuth/API| D[Salesforce]
E[Monitoring] --> B
F[Audit Logs] --> B
G[Access Control] --> B
`}
/>
## Core Security Features
**End-to-end encryption for all data**
* AES-256 encryption for stored credentials
* TLS 1.3 for all API communications
* Encrypted database storage
* Zero-knowledge architecture
**Granular permissions and authentication**
* Role-based access control (RBAC)
* Multi-factor authentication support
* Session management and timeouts
* API key rotation capabilities
**Comprehensive logging and monitoring**
* Complete audit trails for all actions
* Real-time security monitoring
* Anomaly detection and alerting
* Compliance reporting tools
**Privacy-first approach to data**
* Minimal data retention policies
* Right to data deletion
* Privacy by design architecture
* Anonymization capabilities
## Data Handling Principles
Only collect and process data necessary for service functionality
* Purpose limitation for all data collection
* Regular data audits and cleanup
* Automatic data expiration policies
* User control over data sharing
Clear visibility into how your data is used
* Detailed privacy policies
* Real-time data usage dashboards
* Clear consent mechanisms
* Regular transparency reports
Full control over your data and usage
* Granular privacy settings
* Data export capabilities
* Right to deletion
* Consent withdrawal options
## Network Allowlist
**Need to allowlist MeshMesh in your firewall or API gateway?** All sandbox outbound traffic exits through a small fixed set of static IPs. See [Network Allowlist](/docs/network-allowlist) for the current IP list and integration guidance.
## Security Best Practices
**Strong Authentication (Critical)**
Use multi-factor authentication and strong passwords
**Regular Access Reviews (High)**
Periodically review and update user permissions
**Session Management (Medium)**
Log out of sessions when not in use
**Connection Security (Critical)**
Ensure Salesforce API credentials are properly secured
**Data Classification (High)**
Classify data based on sensitivity levels
**Backup Verification (Medium)**
Regularly verify data backup integrity
**Network Security (High)**
Use secure networks for accessing MeshMesh
**Incident Response (High)**
Have a plan for security incident response
**Regular Updates (Medium)**
Keep systems and browsers updated
## Threat Protection
**Protection:** Multi-layered encryption and access controls
* Zero-trust security architecture
* Continuous security monitoring
* Incident response procedures
* Regular penetration testing
**Protection:** Strong authentication and authorization
* Multi-factor authentication
* Role-based access control
* Session monitoring
* Anomaly detection
**Protection:** Comprehensive backup and recovery
* Automated backup processes
* Geographic data replication
* Point-in-time recovery
* Disaster recovery testing
**Protection:** Built-in compliance frameworks
* Automated compliance checking
* Regular compliance audits
* Policy enforcement engines
* Compliance reporting tools
## Incident Response
### Detection
24/7 monitoring and automated alerts for potential security incidents
### Assessment
Rapid impact analysis and classification of the incident severity
### Containment
Immediate threat isolation and mitigation to prevent spread
### Recovery
System restoration and notification to affected customers
**Customer Communication:**
* Immediate notification for incidents affecting data
* Regular updates throughout incident resolution
* Post-incident reports with lessons learned
* 72-hour breach notification for GDPR compliance
## Security Resources
Need help with security questions or want to report a vulnerability?
**Security Issues:** [security@meshmesh.io](mailto:security@meshmesh.io)
**Privacy Questions:** [privacy@meshmesh.io](mailto:privacy@meshmesh.io)
We encourage responsible disclosure of security vulnerabilities. Our security team is available 24/7 to respond to critical security incidents.
# Use Cases
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
import { Steps, Step } from 'fumadocs-ui/components/steps';
Discover how MeshMesh transforms complex Salesforce operations into simple conversations. From Marketing Cloud campaigns to Sales Cloud pipelines, from Service Cloud cases to Data Cloud insights - see what's possible with AI-powered Salesforce management.
**Natural Language Power:** All these use cases work through simple conversation. No technical expertise required - just describe what you want to achieve in plain English.
## Video Walkthroughs
## Industry-Specific Scenarios
} title="E-commerce">
**Retail & Online Shopping**
* Abandoned cart recovery sequences
* Post-purchase follow-up campaigns
* Product recommendation engines
* Inventory-based email triggers
} title="SaaS">
**Software as a Service**
* User onboarding workflows
* Feature adoption campaigns
* Churn prevention strategies
* Usage-based communications
} title="Financial Services">
**Banking & Finance**
* Compliance reporting automation
* Customer lifecycle management
* Fraud alert communications
* Investment opportunity targeting
} title="Healthcare">
**Medical & Health Services**
* Appointment reminder systems
* Health education campaigns
* Patient engagement tracking
* Medication adherence programs
## Cross-Cloud Data Management
} title="Unified Customer Analytics">
**Analyze customer data across Sales, Service, and Marketing Cloud**
**Example Request:**
```
Show me customers who are high-value in Sales Cloud, have recent service cases,
but haven't engaged with marketing emails
```
**Key Benefits:** 360-degree customer view, cross-team insights, better customer experiences
} title="Data Synchronization">
**Keep customer data synchronized across all Salesforce products**
**Example Request:**
```
Update contact preferences in Marketing Cloud when a customer changes
communication settings in Service Cloud
```
**Key Benefits:** Consistent data, reduced manual work, better compliance
} title="Customer Journey Mapping">
**Track complete customer journeys across all touchpoints**
**Example Request:**
```
Map the journey from lead in Sales Cloud to customer in Service Cloud
to marketing subscriber
```
**Key Benefits:** Complete visibility, journey optimization, experience improvement
## Sales & Pipeline Management
} title="Lead Qualification & Routing">
**Automatically qualify leads and route to the right sales reps**
**Example Request:**
```
Show me high-scoring leads from this week and create follow-up tasks
for the appropriate sales reps
```
**Key Benefits:** Faster lead response, better conversion rates, optimized sales process
} title="Pipeline Analysis & Forecasting">
**Analyze sales pipeline and predict revenue outcomes**
**Example Request:**
```
Generate a quarterly forecast report showing probability-weighted revenue
by sales rep and region
```
**Key Benefits:** Accurate forecasting, pipeline visibility, strategic planning
} title="Cross-Cloud Campaign Attribution">
**Track marketing campaign impact on sales outcomes**
**Example Request:**
```
Show me which marketing campaigns generated the most qualified leads
that converted to closed-won opportunities
```
**Key Benefits:** Marketing ROI proof, campaign optimization, budget allocation
## Automation & Workflows
} title="Journey Optimization">
**Analyze and improve customer journey performance**
**Example Request:**
```
Show me where customers are dropping off in my welcome journey
```
**Key Benefits:** Reduced drop-offs, better experiences, higher completion rates
} title="Triggered Email Setup">
**Create and manage triggered email campaigns**
**Example Request:**
```
Set up an abandoned cart reminder sequence with personalized product details
```
**Key Benefits:** Automated engagement, recovery campaigns, timely communications
} title="Workflow Monitoring">
**Monitor and troubleshoot automation workflows**
**Example Request:**
```
Check the status of all automations and show any that failed in the last 24 hours
```
**Key Benefits:** Proactive monitoring, quick issue resolution, reliable automation
## Reporting & Analytics
} title="Performance Dashboards">
**Create comprehensive performance reports**
**Example Request:**
```
Generate a monthly email performance report with opens, clicks, and conversions by segment
```
**Key Benefits:** Clear insights, stakeholder reporting, performance tracking
} title="ROI Analysis">
**Calculate and track marketing ROI**
**Example Request:**
```
Calculate the ROI of my holiday email campaign including revenue attribution
```
**Key Benefits:** Budget justification, campaign optimization, strategic planning
} title="Compliance Monitoring">
**Track and ensure compliance with regulations**
**Example Request:**
```
Show me all unsubscribe requests from the last month and verify they were processed
```
**Key Benefits:** Regulatory compliance, risk mitigation, trust building
## Reusable Workflows
B[Select Approach]
B --> C[Customize for Your Needs]
C --> D[Preview & Test]
D --> E{Works Well?}
E -->|No| C
E -->|Yes| F[Execute Workflow]
F --> G[Save for Future Use]
G --> H[Share with Team]
`}
/>
## Common Solutions
} title="Email Campaigns">
**Pre-built email campaign workflows**
* Welcome email series
* Abandoned cart recovery
* Post-purchase follow-ups
* Re-engagement campaigns
* Event invitations
* Newsletter automation
**Example:** "Set up a welcome series for new subscribers"
} title="Data Operations">
**Data management and analysis workflows**
* Subscriber exports with engagement metrics
* Segmentation queries
* Data cleanup scripts
* Performance analytics
* Compliance reporting
* Journey analytics
**Example:** "Export subscribers who opened emails in the last 30 days"
} title="API Operations">
**Automated API-based workflows**
* Data extension backup and restore
* Subscriber management
* Email send tracking
* Journey status checks
* Asset management
* Automation control
**Example:** "Back up all data extensions to secure storage"
} title="Automation Workflows">
**Journey and automation patterns**
* Welcome journey flows
* Birthday campaigns
* Win-back sequences
* Event-triggered workflows
* Scheduled reporting
* Data synchronization
**Example:** "Create a birthday campaign that sends 3 days before each subscriber's birthday"
## Frequently Used Workflows
} title="Welcome Email Series">
**Multi-step welcome email automation with personalization**
Ask: "Set up a 5-email welcome series for new subscribers with progressive personalization"
**Most Used:** 1,240 times | **Rating:** ⭐ 4.8
} title="Subscriber Export Query">
**Export subscriber data with engagement metrics**
Ask: "Export all subscribers with open rates, click rates, and last engagement date"
**Most Used:** 890 times | **Rating:** ⭐ 4.9
} title="Abandoned Cart Recovery">
**Automated cart abandonment email sequence**
Ask: "Create an abandoned cart sequence: 1 hour, 24 hours, and 3 days after abandonment"
**Most Used:** 756 times | **Rating:** ⭐ 4.7
} title="Data Extension Backup">
**Complete backup and restore for data extensions**
Ask: "Back up all my data extensions and schedule weekly backups"
**Most Used:** 623 times | **Rating:** ⭐ 4.6
## How to Execute Workflows
**Describe Your Goal**
Simply describe what you want to accomplish in plain English. MeshMesh will understand your requirements and suggest the best approach.
**Customize Parameters**
Adjust timing, content, audience segments, and other parameters to match your specific needs. The AI assistant guides you through customization.
**Execute & Monitor**
Run the workflow directly or save it for future use. All workflows integrate seamlessly with your Salesforce connection and can be monitored in real-time.
## Working with Workflows
**Natural Conversation Examples**
**Execute a common workflow:**
```text
Set up a welcome email series for new subscribers from last week
→ Creates and configures the welcome series automatically
```
**Customize for your needs:**
```text
Modify the subscriber export to include engagement data from last 30 days
and filter for customers in California
→ Adjusts parameters to match your specific requirements
```
## Saving & Sharing Workflows
} title="Save Successful Workflows">
**Turn any workflow into a reusable solution**
After completing a task, save it for future use. Perfect for recurring operations or complex multi-step processes.
**Example:** "Save this workflow as 'Monthly VIP Report'"
} title="Add Custom Parameters">
**Make workflows flexible with variables**
Include placeholder values that can be customized each time you run the workflow.
**Example:** "Create a segment query for \{region} customers who purchased \{product\_category}"
} title="Share with Team">
**Collaborate across your organization**
Share workflows with team members to standardize processes and ensure consistency across your Salesforce operations.
**Example:** "Share my 'Campaign Setup' workflow with the marketing team"
## Best Practices
} title="Organization">
* Use clear, descriptive names
* Document purpose and use cases
* Tag by category or department
* Keep workflows focused on specific goals
} title="Testing & Refinement">
* Test workflows before sharing
* Include example values
* Document required permissions
* Track which workflows perform best
} title="Team Collaboration">
* Follow naming conventions
* Include error handling steps
* Add usage examples
* Update regularly based on feedback
**Workflow Management**
* "Show me all my saved workflows"
* "Edit the 'Monthly Report' workflow"
* "Delete the 'Old Campaign' workflow"
* "Duplicate 'Welcome Series' as 'VIP Welcome Series'"
## Complete Workflow Example
**E-commerce: Re-engagement Campaign**
A complete workflow from data analysis to campaign execution
**Identify Inactive Customers**
```
Show me customers who haven't made a purchase in 60 days but were active before that
```
**Analyze Their Preferences**
```
What were the most popular product categories for these inactive customers?
```
**Create Personalized Content**
```
Generate a re-engagement email with personalized product recommendations and a 20% discount
```
**Set Up Automation**
```
Create a follow-up sequence for customers who don't engage with the first email
```
**Monitor and Optimize**
```
Track the campaign performance and show me which product categories are driving re-engagement
```
## Getting Started
} title="Start Simple">
**Begin with basic questions about your data**
Ask "What email campaigns did I send last month?" or "Show me my top performing emails."
Perfect for learning the system and building confidence.
} title="Build Complexity">
**Combine multiple operations once comfortable**
"Find high-value customers who haven't engaged recently and create a win-back campaign."
Layer multiple data sources and actions together.
} title="Automate Everything">
**Move to full automation workflows**
"Set up a workflow that automatically segments customers and sends personalized emails based on their behavior."
Let MeshMesh handle recurring tasks automatically.
***
**Ready to Get Started?**
MeshMesh adapts to your industry, business, data structure, and business goals. Simply describe what you want to achieve, and get step-by-step guidance tailored to your needs.
## Complete Use Case Workflow
B[Describe in MeshMesh]
B --> C[MeshMesh Analyzes Request]
C --> D{Multi-Cloud Data?}
D -->|Yes| E[Query Across Products]
D -->|No| F[Query Single Product]
E --> G[Combine & Analyze Data]
F --> G
G --> H[Generate Insights]
H --> I[Execute Actions]
I --> J[Monitor & Optimize]
`}
/>
# Endpoints
import { Callout } from 'fumadocs-ui/components/callout';
Every endpoint is `GET`, authenticated with a `mesh_sk_` Bearer key, and wrapped in the standard `{ data, meta, next_cursor }` [envelope](/api-reference). All accept the shared `start` / `end` timeframe params; if omitted, they default to the **last 90 days**. `meta.start` / `meta.end` always reflect the window actually queried. The four `/analytics/*` endpoints need the `analytics:read` scope; `/usage` needs `usage:read`.
***
## GET /users
Per-user analytics rows, one per user in your workspace. **Cursor-paginated** — follow `next_cursor` until it is `null`.
### Query parameters
| Param | Type | Default | Description |
| -------- | --------------- | ------------ | -------------------------------------------- |
| `start` | ISO date | last 90 days | Start of the timeframe. |
| `end` | ISO date | last 90 days | End of the timeframe. |
| `cursor` | string | — | Opaque cursor from a previous `next_cursor`. |
| `limit` | integer (1–200) | `50` | Max rows per page. |
### Example response
```json
{
"data": [
{
"userId": "9c1f…",
"email": "ada@example.com",
"name": "Ada Lovelace",
"segment": "active",
"firstLoginDate": "2026-03-02T09:14:00.000Z",
"lastActivityDate": "2026-05-24T17:41:00.000Z",
"firstTaskDate": "2026-03-02T10:01:00.000Z",
"totalCreditsUsed": 412.5,
"inputTokens": 1840221,
"outputTokens": 290114,
"agentCount": 7,
"taskCount": 31,
"uniqueTaskTypes": 4,
"connectionCount": 2,
"role": "member",
"joinedAt": "2026-03-01T00:00:00.000Z"
}
],
"meta": { "start": "2026-03-01", "end": "2026-05-25", "has_more": true },
"next_cursor": "MjAyNi0wMy0wMXw5YzFm"
}
```
`segment` is one of `active`, `lapsed`, or `inactive`. `name`, `firstLoginDate`, `lastActivityDate`, `firstTaskDate`, and `role` may be `null`.
***
## GET /overview
Workspace totals, the user-segment distribution, and the onboarding funnel for the timeframe.
### Query parameters
| Param | Type | Default | Description |
| ------- | -------- | ------------ | ----------------------- |
| `start` | ISO date | last 90 days | Start of the timeframe. |
| `end` | ISO date | last 90 days | End of the timeframe. |
### Example response
```json
{
"data": {
"summary": {
"totalUsers": 48,
"activeUsers": 31,
"lapsedUsers": 9,
"inactiveUsers": 8,
"totalCreditsUsed": 12840.75,
"avgCreditsPerUser": 267.52,
"avgAgentsPerUser": 5.4,
"avgTasksPerUser": 22.1
},
"segmentDistribution": [
{ "segment": "active", "count": 31, "percentage": 65 },
{ "segment": "lapsed", "count": 9, "percentage": 19 },
{ "segment": "inactive", "count": 8, "percentage": 17 }
],
"funnel": [
{ "stage": "total", "label": "Total Users", "count": 48, "percentage": 100 },
{ "stage": "tier-0", "label": "1+ tasks", "count": 40, "percentage": 83 },
{ "stage": "tier-1", "label": "5+ tasks", "count": 28, "percentage": 58 },
{ "stage": "tier-2", "label": "10+ tasks", "count": 19, "percentage": 40 },
{ "stage": "tier-3", "label": "25+ tasks", "count": 9, "percentage": 19 },
{ "stage": "tier-4", "label": "50+ tasks", "count": 3, "percentage": 6 }
]
},
"meta": { "start": "2026-03-01", "end": "2026-05-25" },
"next_cursor": null
}
```
`segment` is one of `active`, `lapsed`, or `inactive`. `segmentDistribution[].percentage` and `funnel[].percentage` are integers (rounded). Funnel stages are `total` (Total Users) followed by task-count tiers `tier-0` (1+ tasks), `tier-1` (5+), `tier-2` (10+), `tier-3` (25+), `tier-4` (50+), and `tier-5` (100+); only stages with a non-zero count are included.
***
## GET /credits
Credit-burn timeseries plus a per-user credit leaderboard.
### Query parameters
| Param | Type | Default | Description |
| ------------- | -------------------------- | ------------ | ------------------------------------ |
| `start` | ISO date | last 90 days | Start of the timeframe. |
| `end` | ISO date | last 90 days | End of the timeframe. |
| `granularity` | `day` \| `week` \| `month` | `day` | Bucket size for the burn timeseries. |
### Example response
```json
{
"data": {
"granularity": "day",
"creditBurn": [
{ "period": "2026-05-23", "totalCredits": 218.4, "userCount": 12 },
{ "period": "2026-05-24", "totalCredits": 301.9, "userCount": 15 }
],
"leaderboard": [
{ "userId": "9c1f…", "email": "ada@example.com", "name": "Ada Lovelace", "totalCredits": 412.5 },
{ "userId": "3b7d…", "email": "alan@example.com", "name": null, "totalCredits": 388.0 }
]
},
"meta": { "start": "2026-03-01", "end": "2026-05-25" },
"next_cursor": null
}
```
`leaderboard[].name` may be `null`.
***
## GET /engagement
Daily/weekly active-user counts, weekly cohort retention, and the weekly user-segment trend (active / lapsed / inactive).
### Query parameters
| Param | Type | Default | Description |
| ------- | -------- | ------------ | ----------------------- |
| `start` | ISO date | last 90 days | Start of the timeframe. |
| `end` | ISO date | last 90 days | End of the timeframe. |
The `dauWau` series respects `start`/`end`. `cohortRetention` and `segmentTrend` are always computed over all-time data — they are not windowed by `start`/`end`.
### Example response
```json
{
"data": {
"dauWau": [
{ "date": "2026-05-23", "dau": 12, "wau": 28 },
{ "date": "2026-05-24", "dau": 15, "wau": 29 }
],
"cohortRetention": [
{ "cohortWeek": "2026-03-02", "cohortSize": 11, "retention": [100, 82, 64, 55] },
{ "cohortWeek": "2026-03-09", "cohortSize": 8, "retention": [100, 75, 50] }
],
"segmentTrend": [
{ "week": "2026-05-04", "active": 18, "lapsed": 4, "inactive": 2 },
{ "week": "2026-05-11", "active": 21, "lapsed": 3, "inactive": 2 }
]
},
"meta": { "start": "2026-03-01", "end": "2026-05-25" },
"next_cursor": null
}
```
In `cohortRetention`, `retention` is a percentage array indexed by weeks since signup — index `0` is the signup week (always `100`).
In `segmentTrend`, each week classifies every member who had joined by then: `active` (used the product that week), `lapsed` (active in an earlier week but not this one), or `inactive` (never active up to that week).
***
## GET /usage
Raw per-call usage records for your workspace — one row per metered LLM call, including token counts, **provider cost**, and credits charged. Tenant-wide: rows for every user are returned, each tagged with `userId` / `email`. **Cursor-paginated** by `(createdAt, id)` — follow `next_cursor` until it is `null`.
This endpoint requires a key with the **`usage:read`** scope — separate from `analytics:read`, because it exposes per-call provider cost. Grant it when you mint the key.
### Query parameters
| Param | Type | Default | Description |
| ---------- | --------------- | ------------ | ----------------------------------------------------- |
| `start` | ISO date | last 90 days | Start of the timeframe. |
| `end` | ISO date | last 90 days | End of the timeframe. |
| `userId` | string | — | Narrow to one user. Filtering only, not a permission. |
| `model` | string | — | Narrow to one model. |
| `provider` | string | — | Narrow to one provider. |
| `cursor` | string | — | Opaque cursor from a previous `next_cursor`. |
| `limit` | integer (1–200) | `50` | Max rows per page. |
### Example response
```json
{
"data": [
{
"tenantId": "org_9f…",
"tenantName": "Acme Inc",
"userId": "9c1f…",
"email": "ada@example.com",
"model": "claude-opus-4-8",
"provider": "anthropic",
"promptTokens": 1840,
"completionTokens": 290,
"totalTokens": 2130,
"cacheCreationInputTokens": 512,
"cacheReadInputTokens": 1024,
"providerCostUsd": "0.04821500",
"creditsActual": "6.450000",
"taskId": "22a1…",
"createdAt": "2026-05-24T17:41:03.221Z"
}
],
"meta": { "start": "2026-03-01", "end": "2026-05-25", "has_more": true },
"next_cursor": "MjAyNi0wNS0yNFQxNzo0MTowMy4yMjFafDIyYTE"
}
```
`providerCostUsd` and `creditsActual` are decimal **strings** (not numbers) to preserve full precision — parse them with a decimal library if you compute on them. `taskId` may be `null` for non-task calls.
The four Analytics endpoints above report consumption only — credits and tokens, never cost or markup. Per-call **provider cost** is available solely through `GET /usage`, which requires the `usage:read` scope.
# Analytics API
import { Callout } from 'fumadocs-ui/components/callout';
import { Tabs, Tab } from 'fumadocs-ui/components/tabs';
The Analytics API is a **read-only** HTTP API that exposes usage analytics for your workspace — the same per-user activity, credit, and engagement data the console dashboard shows. Every request is scoped to the tenant the API key belongs to; there is no way to read another workspace's data.
Aggregated endpoints live under `https://gateway.meshmesh.io/api/v1/public/analytics`, and the raw per-call feed under `https://gateway.meshmesh.io/api/v1/public/usage`. All respond with JSON.
## Creating a key
API keys are created in the console:
1. Go to **Workspace Admin → API Keys**.
2. Click **Create key**, give it a name, and confirm.
3. Copy the key — it is shown **once** and never again. Store it somewhere secure (a secrets manager, not source control).
If you lose a key you cannot recover it. Create a new one and revoke the old one. Revoking a key takes effect within \~60 seconds.
Keys are minted by workspace admins and carry one or more scopes: `analytics:read` for the aggregated `/analytics/*` endpoints, and `usage:read` for the raw [`/usage`](/api-reference/endpoints#get-usage) feed (which includes per-call provider cost). Pick the scopes when you create the key.
## Authentication
Pass the key as a Bearer token. API keys are prefixed `mesh_sk_`.
```
Authorization: Bearer mesh_sk_…
```
Requests with a missing, invalid, revoked, or expired key get a generic `401`. A key that is valid but lacks the scope an endpoint requires (e.g. `usage:read` for `/usage`) gets a `403`.
```bash
curl https://gateway.meshmesh.io/api/v1/public/analytics/overview \
-H "Authorization: Bearer mesh_sk_your_key_here"
```
```js
const res = await fetch(
'https://gateway.meshmesh.io/api/v1/public/analytics/overview',
{ headers: { Authorization: `Bearer ${process.env.MESH_ANALYTICS_KEY}` } },
);
const { data, meta, next_cursor } = await res.json();
```
## Timeframe
Every endpoint accepts an optional timeframe via two query params:
| Param | Type | Description |
| ------- | -------------- | -------------------------------------------------------------- |
| `start` | ISO date (UTC) | Start of the window. If omitted, defaults to the last 90 days. |
| `end` | ISO date (UTC) | End of the window. If omitted, defaults to the last 90 days. |
When `start`/`end` are omitted, all four endpoints default to the **last 90 days**. `meta.start` and `meta.end` in the response always reflect the window actually queried (as ISO dates), whether you specified it or got the default.
```bash
curl "https://gateway.meshmesh.io/api/v1/public/analytics/credits?start=2026-03-01&end=2026-05-25" \
-H "Authorization: Bearer mesh_sk_…"
```
If `start` is after `end`, or either value isn't a valid ISO date, you get a `400`.
Analytics data is available from approximately **2026-03-01**. Requesting an earlier window returns partial or empty results rather than an error.
## Response envelope
Every endpoint returns the same envelope:
```json
{
"data": { "...": "endpoint-specific payload" },
"meta": { "start": "2026-03-01", "end": "2026-05-25" },
"next_cursor": null
}
```
* **`data`** — the endpoint-specific payload (an array or object; see each endpoint).
* **`meta.start` / `meta.end`** — the window actually queried, echoed back as ISO dates. When you omit `start`/`end`, these reflect the default last-90-days window.
* **`next_cursor`** — an opaque pagination cursor, or `null` when there are no more pages.
### Cursor pagination
Only `GET /users` is paginated. When more rows are available, `next_cursor` is a non-null string and `meta.has_more` is `true`. Pass the value back as the `cursor` query param to fetch the next page; keep going until `next_cursor` is `null`.
```js
let cursor = null;
const rows = [];
do {
const url = new URL('https://gateway.meshmesh.io/api/v1/public/analytics/users');
url.searchParams.set('limit', '200');
if (cursor) url.searchParams.set('cursor', cursor);
const res = await fetch(url, {
headers: { Authorization: `Bearer ${process.env.MESH_ANALYTICS_KEY}` },
});
const body = await res.json();
rows.push(...body.data);
cursor = body.next_cursor;
} while (cursor);
```
`limit` accepts 1–200 and defaults to 50.
## Rate limits
Each key is limited to **120 requests/minute** and **10,000 requests/day** by default. When you exceed a limit you get a `429` with these headers:
| Header | Meaning |
| --------------------- | ------------------------------------- |
| `Retry-After` | Seconds to wait before retrying. |
| `RateLimit-Limit` | Request quota for the current window. |
| `RateLimit-Remaining` | Requests remaining in the window. |
| `RateLimit-Reset` | Seconds until the window resets. |
Back off and retry after `Retry-After` seconds.
## Errors
Errors use [RFC 9457](https://www.rfc-editor.org/rfc/rfc9457) problem documents with the `application/problem+json` content type:
```json
{
"type": "https://docs.meshmesh.io/errors/invalid-request",
"title": "Invalid request parameters",
"status": 400,
"detail": "`start` is not a valid ISO date."
}
```
| Status | When |
| ------ | ----------------------------------------------------------------------- |
| `400` | Malformed query params (bad date, `start` after `end`). |
| `401` | Unauthorized — API key is missing, invalid, revoked, or expired. |
| `403` | Insufficient scope — the key lacks the required `analytics:read` scope. |
| `429` | Rate limit exceeded (see above). |
## Endpoints
See the [Endpoints](/api-reference/endpoints) reference for all available endpoints and their response shapes.
# API Keys (BYOK)
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
Bring Your Own Key lets you use your own AI provider API keys instead of MeshMesh's built-in models. When you add your own key, API requests go directly to your provider account, giving you control over costs and usage tracking.
## Why Use Your Own Keys
Using your own API keys makes sense when you have an existing enterprise agreement with Anthropic or OpenAI and want to use your negotiated pricing. It's also useful if you need detailed usage tracking in your own provider dashboard, or if you want your AI costs consolidated with your other AI usage.
## Adding an Anthropic Key
Get your API key from [console.anthropic.com](https://console.anthropic.com) under API Keys. In MeshMesh, go to Settings → API Keys, select Anthropic, paste your key, and save.
To verify it's working, create a test task and check your Anthropic console to confirm the usage appears. You'll have access to Claude 3.5 Sonnet, Claude 3.5 Haiku, and Claude 3 Opus depending on your account's model access.
## Adding an OpenAI Key
Get your API key from [platform.openai.com](https://platform.openai.com) under API Keys. In MeshMesh, go to Settings → API Keys, select OpenAI, paste your key, and save.
Verify by creating a test task and checking your OpenAI usage dashboard. Available models include GPT-4o, GPT-4 Turbo, and GPT-3.5 Turbo.
## Managing Keys
Your saved keys appear in Settings → API Keys with the key partially masked (only the last 4 characters visible). You can update a key by clicking it and entering a new one, or remove it entirely to switch back to MeshMesh's built-in models.
Removing your BYOK key doesn't affect completed tasks. It just means future tasks will use MeshMesh's models instead of your own.
## Security
API keys are encrypted at rest and never appear in logs or task outputs. When you use BYOK, API calls go directly to the provider — they don't pass through MeshMesh's servers. The desktop app stores keys locally on your machine.
Keep your API keys secret. Anyone with your key can make requests billed to your account.
## Troubleshooting
**"Invalid API Key"** means the key wasn't recognized. Check that you copied it correctly without extra spaces, and verify it hasn't been revoked in your provider dashboard.
**"Rate Limit Exceeded"** means your provider account hit its rate limit. Wait for limits to reset or upgrade your plan.
**"Insufficient Credits"** means your provider account needs payment. Add credits or update billing in your provider dashboard.
Start using your API keys
Connect Salesforce and other services
# Connections
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
import { Tabs, Tab } from 'fumadocs-ui/components/tabs';
import { Accordion, Accordions } from 'fumadocs-ui/components/accordion';
import { Steps } from 'fumadocs-ui/components/steps';
# Connections & Integrations
Connect MeshMesh to your Salesforce products, extend capabilities with MCPs (Model Context Protocol), and enable agent-to-agent communication with A2A. All connections are encrypted and you maintain full control over data access.
Quick Setup: Most connections take less than 5 minutes to configure
## Overview
MeshMesh integrates with your tools and services through three main connection types:
Direct connections to Sales Cloud, Marketing Cloud, Service Cloud, Data Cloud, and more using Salesforce OAuth and APIs
Extend AI capabilities with specialized tools, external services, databases, and custom integrations
Enable communication between AI agents for multi-agent workflows and specialized service delegation
## Salesforce Products
Connect your Salesforce ecosystem to query data, manage campaigns, automate workflows, and more.
### Supported Products
**Salesforce Core Products**
Connect via **Salesforce OAuth** for:
* Sales Cloud
* Revenue Cloud
* Service Cloud
* Industry Cloud
* Data Cloud
* Agentforce
* Einstein
* CRM Analytics
* Commerce Cloud
* Experience Cloud
* Marketing Cloud Growth
* Marketing Cloud Advanced
* Marketing Cloud Next
* Flows
* APEX
* Lightning Apps
**What is Salesforce OAuth?**
A secure authentication method that allows MeshMesh to connect to your Salesforce org through your browser, without requiring credentials to be stored.
**Permissions:** MeshMesh can only act on Salesforce products, features, and data capabilities based on your specific user permissions in Salesforce.
**Marketing Cloud Engagement**
Connect via **Install Packages** with Server-to-Server component for:
* Data Extensions
* Email Studio
* Journey Builder
* Automation Studio
* Content Builder
**Required Information:**
* Client ID and Secret
* Authentication Base URI
* REST Base URI
* SOAP Endpoint URL
**Permissions:** Data Extensions, Email, Journeys, Automation (Read/Write)
**Learn more:** [Marketing Cloud Install Packages](https://developer.salesforce.com/docs/marketing/marketing-cloud/guide/install-packages.html)
**Mulesoft MCP Gateway**
Connect via **MCP Server Integration** for:
* API-led connectivity
* Integration patterns
* MCP server capabilities
**Required Information:**
* MCP Gateway access credentials
* API endpoint configuration
* Integration permissions
**Learn more:** [Mulesoft MCP Server](https://docs.mulesoft.com/mulesoft-mcp-server/)
### Salesforce Setup
### Select Salesforce from Connections
Navigate to the Connections page and select "Salesforce"
### Login to your Salesforce Org through the browser
You'll be redirected to Salesforce to authenticate with your credentials
### You're connected!
Your Salesforce org is now connected and ready to use
You can add multiple Salesforce orgs by repeating this process
**Important:** MeshMesh can only act on Salesforce products, features, and data capabilities based on your specific permissions to work in Salesforce. If something in Salesforce is not acting as expected through MeshMesh, a common culprit is your user permissions to administrate Salesforce effectively.
### Create Installed Package
1. Navigate to Setup → Apps → Installed Packages
2. Create new package with Server-to-Server component
3. Configure API permissions:
* Data Extensions (Read/Write)
* Email (Read/Write)
* Journeys (Read/Write)
* Automation (Read/Write)
### Save Credentials
Copy and securely store:
* Client ID
* Client Secret
* Authentication Base URI
* REST Base URI
* SOAP Endpoint URL
### Connect in MeshMesh
1. Go to Connections page
2. Click "Add New Connection"
3. Select "Marketing Cloud Engagement"
4. Enter your credentials
5. Test connection and save
**Reference:** [Install Packages Documentation](https://developer.salesforce.com/docs/marketing/marketing-cloud/guide/install-packages.html)
### Configure Gateway Access
1. Set up MCP Gateway credentials
2. Configure API endpoint
3. Enable integration permissions
### Test Gateway Connectivity
Verify your MCP Gateway is accessible and responding
### Connect in MeshMesh
1. Go to Connections page
2. Click "Add New Connection"
3. Select "Mulesoft MCP Gateway"
4. Enter gateway credentials
5. Test connection and save
**Reference:** [Mulesoft MCP Server Documentation](https://docs.mulesoft.com/mulesoft-mcp-server/)
## MCPs (Model Context Protocol)
MCPs extend your AI tasks with specialized capabilities—from querying databases to calling APIs to generating content. Think of MCPs as plugins that give your AI superpowers.
### What are MCPs?
Model Context Protocol is a standardized way for AI models to interact with external systems. MCPs enable your tasks to:
* Access real-time data from external services
* Query databases and data warehouses
* Call third-party APIs
* Generate specialized content
* Communicate with other AI agents
### Built-in MCPs
Several MCPs are automatically available with no additional setup required:
Query Data Extensions, manage Journeys, create Email campaigns, and automate marketing workflows
Connect to and communicate with other AI agents using the Agent-to-Agent protocol
Generate structured content, documents, and formatted outputs
Create platform-specific content for LinkedIn, Twitter, Facebook, and more
Built-in MCPs are ready to use immediately in any task—just reference them naturally!
**Example Usage:**
```
"Query Marketing Cloud for all active journeys"
"Generate a LinkedIn post about our new product"
"Call the weather-agent to get today's forecast"
"Create a summary document from this data"
```
### Import Custom MCPs
Add any standard MCP to extend capabilities further:
### Go to Connections Page
Navigate to Connections → Import MCP Config
### Paste MCP Configuration
Enter your MCP JSON configuration:
```json
{
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-brave-search"],
"env": {
"BRAVE_API_KEY": "YOUR_API_KEY_HERE"
}
}
```
**Other examples:**
```json
{
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-github"],
"env": {
"GITHUB_TOKEN": "ghp_your_token_here"
}
}
```
```json
{
"command": "node",
"args": ["/path/to/your/mcp-server.js"],
"env": {
"DATABASE_URL": "postgresql://...",
"API_KEY": "your-key"
}
}
```
```json
{
"url": "https://api.example.com/mcp",
"headers": {
"Authorization": "Bearer YOUR_TOKEN",
"Content-Type": "application/json"
}
}
```
### Auto-Detection
MeshMesh automatically detects:
* Required credentials and environment variables
* Field types (passwords, URLs, numbers, text)
* MCP communication type (stdio, HTTP, SDK)
### Enter Credentials
Fill in the detected credential fields securely
### Save and Use
Your MCP is now available in all tasks!
Any MCP following the standard Model Context Protocol will work with MeshMesh
### Using MCPs in Tasks
Once configured, MCPs work seamlessly with natural language. The AI automatically selects the right MCP based on your request:
```
"Show me all Data Extensions with more than 10,000 subscribers"
"Create a new journey for abandoned cart emails"
"List all emails sent in the last 7 days with open rates"
"Update the welcome journey to include a 2-day delay"
```
```
"Search for recent articles about AI agents using Brave Search"
"Query the customer database for accounts created this month"
"Get the latest stock price for TSLA from the finance API"
"Fetch weather forecast for San Francisco"
```
```
"Create a social media post for LinkedIn about our new feature"
"Generate a summary document from this meeting transcript"
"Write a tweet thread about today's product launch"
"Create an artifact with Q4 sales data formatted as a table"
```
## A2A (Agent-to-Agent Protocol)
A2A enables AI agents to discover and communicate with each other, allowing for sophisticated multi-agent workflows and specialized service delegation.
### What is A2A?
The Agent-to-Agent protocol allows different AI agents to:
* **Discover** each other's capabilities automatically
* **Communicate** using standardized message formats
* **Collaborate** on complex, multi-step tasks
* **Delegate** specialized work to expert agents
The A2A MCP is built-in and available immediately—no manual configuration required!
### Use Cases
Connect to domain-specific agents like weather services, financial analysis, data analytics, or industry-specific tools
Orchestrate complex tasks where different agents handle their areas of expertise (research, analysis, content creation)
Bridge MeshMesh with A2A-compatible services from other platforms and vendors
Build modular AI systems where each agent has clear responsibilities and can be updated independently
### Getting Started with A2A
**Basic A2A Operations:**
Add an A2A-compatible agent to your session:
```
"Register the weather agent at https://weather.example.com/a2a"
"Add the analytics-agent from https://analytics-api.company.com"
```
Agents are stored in an in-memory registry (session-scoped) for security.
List registered agents and their capabilities:
```
"List all registered A2A agents"
"What can the weather-agent do?"
"Show me the capabilities of the analytics-agent"
```
Send requests to specific agents:
```
"Ask the weather-agent for the forecast in Seattle"
"Have the analytics-agent summarize Q4 sales data"
"Call the translation-agent to convert this text to Spanish"
"Get market analysis from the finance-agent for TSLA"
```
The AI handles the A2A protocol communication automatically.
### A2A Message Flow
>MeshMesh: "Call weather-agent for Seattle"
MeshMesh->>A2A_MCP: Parse request
A2A_MCP->>External_Agent: A2A Protocol Request
Note over A2A_MCP,External_Agent: Standardized message format
External_Agent->>A2A_MCP: Response (weather data)
A2A_MCP->>MeshMesh: Formatted result
MeshMesh->>User: "Seattle: 72°F, Partly cloudy..."
`}
/>
### A2A Security Features
* **In-memory registry** - Agents registered per session (not persisted)
* **Network timeouts** - Automatic timeout handling for unresponsive agents
* **Error handling** - Graceful degradation when agents are unavailable
* **Future authentication** - OAuth2/JWT support planned
A2A agents are session-scoped for security—they don't persist between sessions and must be re-registered
## Connection Architecture
All connection types work together seamlessly in your AI tasks:
MeshMesh[MeshMesh AI]
MeshMesh --> SF[Salesforce Connections]
MeshMesh --> MCP[MCP Integrations]
MeshMesh --> A2A[A2A Agents]
SF --> SF1[Sales Cloud via OAuth]
SF --> SF2[Marketing Cloud via API]
SF --> SF3[Mulesoft via MCP Gateway]
MCP --> MCP1[Built-in MCPs SFMC, A2A, etc.]
MCP --> MCP2[Custom MCPs External Services]
MCP --> MCP3[Community MCPs GitHub, DBs, etc.]
A2A --> A2A1[Registered Agents Session-scoped]
A2A --> A2A2[External Services A2A-compatible]
SF1 --> Data[(Your Data)]
SF2 --> Data
SF3 --> Data
SF4 --> Data
MCP2 --> Data
MCP3 --> Data
A2A1 --> Data
A2A2 --> Data
`}
/>
## Security
All credentials encrypted at rest and in transit with zero-knowledge architecture and SOC 2 compliance
You control permissions through Salesforce scopes, MCP configurations, and connection settings. Audit logs track all actions.
Credentials never exposed in logs, task outputs, or error messages
Automatic token refresh keeps connections active without manual intervention
Enable, disable, or revoke any connection instantly without affecting others
A2A agents are session-scoped and don't persist, reducing security risks
**For all connections:**
* Never share API credentials or tokens publicly
* Use least-privilege permissions and scopes
* Regularly rotate credentials and API keys
* Monitor connection audit logs
* Review permissions before enabling new connections
**For Salesforce:**
* Regularly review and audit connected apps
* Use dedicated users with appropriate permissions for integration work
**For MCPs:**
* Never commit MCP configurations with real credentials to version control
* Use environment variables for sensitive data
* Review MCP permissions before importing
**For A2A:**
* Only register trusted A2A agents
* Verify agent URLs before registration
* Monitor agent calls in audit logs
## Troubleshooting
**Symptoms:** Authentication errors or permission denied
**Solutions:**
* Ensure you're logging in with the correct Salesforce org credentials
* Check that your user has the necessary permissions in Salesforce
* Try disconnecting and reconnecting your Salesforce org
* Verify your Salesforce org is active and accessible
* Clear browser cache and cookies, then try again
**Symptoms:** Invalid credentials or permission errors
**Solutions:**
* Verify Client ID and Secret are correct
* Check that Authentication Base URI matches your MC instance
* Ensure Installed Package has required permissions enabled
* Confirm package has Server-to-Server component
* Test credentials directly via Postman or API client
**Symptoms:** Task hangs or times out when using an MCP
**Solutions:**
* Verify the MCP service/script is running
* Check network connectivity for HTTP-based MCPs
* Review MCP logs in Connections page
* Ensure credentials are valid and not expired
* Try disabling and re-enabling the MCP
* For custom MCPs, verify command/script paths
**Symptoms:** Error when importing MCP configuration
**Solutions:**
* Validate JSON syntax using a JSON validator
* Ensure all required fields are present (command or url)
* Check that command paths are absolute or in PATH
* Verify node/npm/npx is installed for npm packages
* Review error message for specific field issues
**Symptoms:** "Agent not found" or connection timeout
**Solutions:**
* Verify the agent URL is correct and accessible
* Check that the agent supports A2A protocol
* Ensure the agent service is running and responsive
* Try re-registering the agent in your current session
* Review network timeouts and agent health status
**Symptoms:** "Authentication failed" after previously working
**Solutions:**
* Wait for automatic token refresh (usually happens automatically)
* For Salesforce: Try re-authenticating through the browser OAuth flow
* For Marketing Cloud: Verify Client ID/Secret are still valid
* For custom MCPs: Check if credentials need rotation
* If persistent, try reconnecting from scratch
**Symptoms:** "Access denied" or "Insufficient privileges"
**Solutions:**
* **Salesforce Core:** Check user scopes and permissions
* **Marketing Cloud:** Verify Installed Package permissions
* **MCPs:** Review MCP configuration permissions
* Contact your admin to grant required permissions
### Build Your Own
Want to create custom MCPs or A2A agents?
* [MCP Specification](https://modelcontextprotocol.io/introduction)
* [MCP SDK Documentation](https://modelcontextprotocol.io/docs/sdk)
* [Example MCP Servers](https://github.com/modelcontextprotocol/servers)
* [A2A Protocol Specification](https://a2a-protocol.io)
Any tool following the MCP or A2A standards will work with MeshMesh—no proprietary integration needed!
## Next Steps
Start using your connections in AI tasks
Use MeshMesh with natural language
Learn effective AI communication
See real-world usage examples
# Agents
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
An **agent** in MeshMesh is an AI assistant designed to accomplish a specific goal or objective. While you can use MeshMesh for quick questions, agents are really built to execute work for you — implementing a Sales Cloud configuration, building an Agentforce assistant, analyzing your pipeline, or cleaning up data across your org.
When you create an agent, you're giving it a mission. The agent figures out the steps needed, connects to your Salesforce org, and works through the objective. You can watch its progress, provide guidance when it asks questions, and review what it produces.
## Plan Mode vs Build Mode
MeshMesh offers two ways to work depending on how much oversight you want.
**Plan Mode** gives the agent read-only access to your connections. It can analyze your org, access the web for research, and create a detailed plan — but it won't make any changes until you switch to Build Mode. You see exactly what it intends to do, review the plan, ask for modifications, and then decide whether to proceed.
**Build Mode** lets the agent make changes to your org. The agent starts executing as soon as you describe what you want — faster for straightforward tasks when you already know what you need.
You can switch between modes anytime from the mode selector.
## Skills
Agents have deep expertise in Salesforce products through **skills**. When an agent uses a skill, it has access to APIs, best practices, and implementation patterns for that product.
| Skill | What It Does |
| ------------------------ | ------------------------------------------------------------------------ |
| **Data Cloud** | Data streams, segments, activation targets, identity resolution, queries |
| **Agentforce** | AI agents, GenAI functions, plugins, planners, deployment |
| **Marketing Cloud** | Journeys, emails, data extensions, automations, transactional messaging |
| **Marketing Cloud Next** | MC Next segments, bulk contact ingestion, identity resolution |
| **Metadata** | Apex, Flows, LWC, custom objects, profiles, permission sets |
The agent automatically selects the right skill based on your request. You can also explicitly ask for one.
Learn how to work with each skill in detail
## Connected Systems
Through [integrations](/docs/concepts/integrations), agents can connect to:
* **Salesforce orgs** — Read and deploy metadata, query data, manage configurations
* **Marketing Cloud instances** — Content Builder, journeys, automations, data extensions
* **Data Cloud** — Data streams, segments, identity resolution, activation targets
## Artifacts and References
Agents work with two types of content:
**Artifacts** are outputs agents create — charts, reports, emails, tables, documents. These are the deliverables you asked for or outputs generated along the way. [Learn more about artifacts](/docs/concepts/artifacts).
**References** are inputs you provide — brand kits, requirements, data files, examples. These help agents understand your context and produce better results. [Learn more about references](/docs/concepts/references).
## Next Steps
Get started with agents in MeshMesh
Learn how to leverage product-specific expertise
# Artifacts
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
**Artifacts** are outputs created by agents during execution. They're the deliverables — the things you asked for or the things the agent produces along the way.
## Types of Artifacts
| Type | Examples |
| ------------- | ------------------------------------------------ |
| **Charts** | Line, bar, pie, funnel, scatter plots |
| **Tables** | Data tables, comparison tables, pricing grids |
| **Reports** | Executive summaries, performance reports, audits |
| **Emails** | HTML emails, text emails, templates |
| **Pages** | Landing pages |
| **Documents** | Plans, specs, documentation |
**Reports** — Analysis and insights the agent generates. Pipeline reports, audit findings, data analysis, performance summaries. Reports help you understand what the agent found and what it recommends.
**Generated assets** — HTML artifacts like emails, landing pages, and other content the agent creates for you. These are ready to use or export.
## Requesting Artifacts
Just ask for what you need:
```
Create a chart showing email performance over the last 90 days
```
```
Build a report summarizing Q4 campaign performance
```
```
Create a welcome email with our brand colors
```
## Updating Artifacts
Don't like something? Ask for changes:
```
Update that chart to show the last 90 days instead of 30
```
```
Change the email subject line to something more urgent
```
```
Make the table sortable by any column
```
The agent modifies the existing artifact rather than creating a new one.
## Using Your Brand Kit
If you have a [brand kit](/docs/guides/brand-kits) configured, the agent applies it automatically to visual outputs:
```
Create an email using my brand kit
```
The agent uses your colors, fonts, and logo. You can also be specific:
```
Create a report using our brand colors but with a more
minimal design than usual
```
## Saving Artifacts as References
When the agent creates something useful, save it for reuse:
```
Save this email template as a reference called "Welcome Series v2"
```
```
Save this report format as a reference for monthly reporting
```
Saved references can be used across multiple tasks.
## Exporting Artifacts
Download artifacts to use outside MeshMesh:
```
Export this report as a PDF
```
```
Download the email HTML
```
```
Export this chart as an image
```
For complex reports you'll run repeatedly, ask the agent to create a blueprint so you can regenerate it with fresh data anytime.
## Next Steps
Understand how agents create artifacts
Start producing artifacts
# Blueprints
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
**Blueprints** are pre-built templates that give your agent a head start. Each blueprint is optimized for a specific type of work and includes context the agent uses to produce better results.
## Example Blueprints
**Plan & Build a Sales Cloud Implementation** — Configure leads, accounts, contacts, opportunities, sales processes, and pipeline dashboards.
**Analyze My Sales Pipeline** — Get insights on pipeline coverage, deal velocity, at-risk opportunities, and forecast accuracy.
**Tech Debt Audit of Salesforce Core** — Review custom code, object model, automation, and security configurations.
## Using Blueprints
To use a blueprint, select it when creating a new agent. The blueprint provides a starting point — you customize it with your specific requirements.
When blueprints don't fit your needs, create an agent from scratch by describing what you want in your own words.
## Next Steps
Get started with blueprints
Add context to help blueprints work better
# Business Context
import { Card, Cards } from 'fumadocs-ui/components/card';
**Business context** is organizational knowledge that helps agents understand how your company works. Business context is currently in early preview. While documents provide specific requirements, business context fills in the background — the processes, terminology, and structures that shape how work gets done.
When you create business context, you select which connections to include. An agent then analyzes your connected systems and automatically creates a business context reference for you — extracting the knowledge it needs to work effectively in your environment.
Just like any other reference, you can ask your agent to make changes to your business context as your organization evolves.
## Next Steps
Learn how to create business context
Understand how all reference types work together
# Context Model
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
The **Cloud Control Context Model** is what enables MeshMesh to accomplish complex Salesforce use cases. It's a continuously learning system with deep knowledge of Salesforce data models, APIs, products, features, and common workflows.
## How It Works
When you describe a goal, the context model provides the AI with everything it needs to navigate Salesforce programmatically:
* **Platform knowledge** — Understanding of how Salesforce products work and interact
* **API expertise** — Knowledge of available endpoints, data structures, and best practices
* **Use case patterns** — Learned approaches for common scenarios
The model learns from successful executions, continuously improving its understanding of what works best.
## Current Version: CC5.1
CC5.1 is our largest context model, with deep understanding across 18+ Salesforce products:
* **Data Cloud** — Data unification, customer profiles, identity resolution, real-time segmentation, web SDK, TypeScript SDK, data graphs, transforms, query API, real-time actions, clean rooms, scoring models
* **Marketing Cloud** — Email, SMS, push messaging, marketing automation, content personalization, Campaign Flows, Distributed Marketing, MobileConnect two-way SMS
* **Customer Journeys** — Journey Builder automation, multi-channel orchestration, trigger and decision logic
* **Agentforce** — Agent Script (.agent files + sf agent CLI), deployment, web deployment, event-log debugging, Builder UI, GenAI Functions, Plugins, Planners, prompt templates, Apex services
* **Commerce Cloud** — E-commerce workflows, product catalogs, order management, inventory across LocationGroups, OMS with Connect API, product visibility and entitlements
* **Sales Cloud & Service Cloud** — CRM data, processes, automation, reporting, Command Center, Next Best Action, social media response patterns
* **Platform** — Apex patterns, browser automation, bulk API, CI/CD, custom fields, metadata deployment, page layouts, permissions, Platform Events, SOQL, approval processes, validation rules, reports
* **Field Service** — Dispatcher Console, Service Resources, Asset Attributes, Parts Consumption
* **Experience Cloud** — CMS workspace permissions, Embedded Messaging security setup, deployment patterns
* **Lightning Apps, Flows, Loyalty Management** — Path Assistant, theming, browser automation, quick promotions
CC5.1 also covers Health Cloud, Education Cloud, Public Sector Solutions, Nonprofit Cloud, Consumer Goods Cloud, OmniStudio, Revenue Cloud / CPQ, CRM Analytics, Tableau Next, Marketing Intelligence, Marketing Cloud Account Engagement (B2B), and Personalization.
## Version Evolution
**CC1** — Initial context model focused on Salesforce Marketing Cloud. Basic email campaign automation and core API understanding.
**CC2** — Expanded to core Salesforce products. Cross-cloud integration patterns, CRM and Service Cloud context.
**CC3** — Deep Data Cloud understanding, Customer Journey orchestration, Agentforce capabilities, Commerce Cloud automation, and advanced multi-cloud workflows.
**CC4** — Production-stable release with refined cross-cloud patterns and improved reliability across existing product coverage.
**CC5** — Major expansion to new clouds including Health Cloud, Education Cloud, Public Sector Solutions, and Nonprofit Cloud. Enhanced platform capabilities.
**CC5.1** — Our largest context model. Coverage across 18+ Salesforce products with substantially deeper Agentforce, Data Cloud, and Platform understanding. Includes Consumer Goods Cloud, OmniStudio, Revenue Cloud / CPQ, CRM Analytics, Tableau Next, and Marketing Intelligence.
## Continuous Learning
The context model improves automatically:
* **Learns from execution** — Each successful agent adds to its knowledge
* **Refines patterns** — Improves understanding of what works best
* **Expands context** — Gains deeper knowledge of Salesforce capabilities
* **Adapts strategies** — Develops better approaches to common challenges
Your MeshMesh installation automatically receives improvements within your version.
## Next Steps
Put the context model to work
See how agents produce outputs
# Integrations
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
**Integrations** are connections between MeshMesh and external systems. They give agents the ability to read data, execute actions, and work across your technology stack.
## Built-in Integrations
MeshMesh includes built-in integrations for:
* **Salesforce Platform** — Connect to production orgs, sandboxes, or custom domains via OAuth
* **Salesforce Marketing Cloud Engagement** — Email campaigns, journeys, data extensions, and automations
* **Mulesoft Anypoint Platform** — API-led connectivity and integration patterns
* **B2C Commerce Cloud** — Code deployment, WebDAV file management, job execution, and sandbox management
* **Image Generation (Gemini)** — Generate images using Google Gemini
* **GitHub** — Repository access and git operations via Personal Access Token
* **Slack** — Send messages, manage channels, and interact with Slack workspaces via Bot Token
## Custom MCP Servers
MeshMesh supports the Model Context Protocol (MCP), so you can connect any MCP-compatible server to extend your agent's capabilities — GitHub, Jira, Braze, and more.
## Next Steps
Set up your Salesforce org connection
Connect Marketing Cloud, Mulesoft, and custom MCPs
# The Library
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
The **Library** is where your workspace's [references](/concepts/references), [skills](/guides/using-skills), and collections live. It's how you find context that others have shared, and how admins curate what the whole workspace can discover.
## Mine and Discover
The Library has two tabs:
* **Mine** — references you own or have starred.
* **Discover** — everything shared with you: references shared directly, plus anything your admins have published workspace-wide.
Workspace **admins** also see a **Review** tab for approving references that members have requested to publish. A **Types** filter and search narrow what you see.
## Collections
**Collections** are curated groups of references — for example, everything an agent needs for a particular product or workflow. You can add an entire collection to an agent or workbench in one action, instead of attaching references one at a time.
## Skills in the Library
[Skills](/guides/using-skills) are a kind of reference, so they live in the Library too. In **Discover**, they appear together as a **Skills** collection.
The Skills collection only appears once your workspace has at least one skill. If you don't see it yet, none have been added.
## Next Steps
Browse, search, and organize with collections
Give agents product expertise
# Recipes
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
**Recipes** turn work you've already done into something repeatable. When an agent completes a task and you like the result, you can ask it to capture the steps as a recipe — a reusable guide that any agent can follow to reproduce the same outcome.
## When to Create a Recipe
* You completed a setup or configuration and want to do the same thing again later
* You want to share a proven workflow with teammates so they don't have to figure it out from scratch
* You built something that works and want to standardize it as a best practice
## How It Works
After completing work with an agent, just ask: *"Turn this into a recipe"*, *"Save this as a runbook"*, or *"Document what we just did."* The agent reviews the conversation, asks a few scoping questions, and creates a recipe you can reuse or share.
Recipes are written for agents, not humans. The instructions are precise and machine-executable, so when you attach a recipe to a new agent, it can follow the steps without ambiguity.
## Discovering Recipes
You can find recipes in the **References** tab — use the sidebar to filter by type, browse shared recipes from your team, or star the ones you use most. Recipes shared with your organization appear in the Shared tab automatically.
## Next Steps
Step-by-step guide to creating, sharing, and using recipes
Learn how references work in MeshMesh
# References
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
**References** are documents, files, and requirements you attach to help the agent understand what you need and how to deliver it.
You might attach a product brief so the agent knows the features to highlight, a brand guide so it matches your tone, or a spreadsheet with data to analyze. The more relevant context you provide, the better the output.
## What to Include
* **Requirements and specs** — What the agent needs to deliver
* **Examples** — Sample emails, reports, or templates to follow
* **Data** — Spreadsheets or CSVs to analyze or use as input
* **Brand and style guides** — Tone of voice, terminology preferences
* **Process documentation** — How things work in your org
## Supported Formats
* **Documents** — `.docx`, `.pdf`, `.txt`, `.md`
* **Spreadsheets** — `.xlsx`, `.csv`
* **Images** — `.png`, `.jpg`, `.gif`
## Next Steps
Learn how to add and manage references
Discover and organize references across your workspace
Start using references in your agents
# Workbenches
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
A **workbench** is a long-lived workspace that hosts one or more [agents](/docs/concepts/agents). Every agent inside shares the same filesystem, connections, and references.
## When to use one
Use a workbench when you want **separate concerns running side by side** instead of one agent juggling everything. A long single-agent run gets slow and expensive; splitting the work across focused agents keeps each conversation tight and lets them run in parallel.
A few realistic shapes:
* One agent owns sales pipeline analysis, another owns marketing performance, an orchestrator agent checks both on a schedule and posts a weekly summary.
* One agent maintains your data model, another runs ad-hoc queries on it — same connections, separate jobs.
* A long-lived "ops" workbench you keep coming back to over weeks: same connections, same references, fresh agents per question.
For a one-off question, a standalone agent is still the right shape.
## Agents and codenames
Each agent in a workbench has its own chat and history but lives in the same sandbox. New agents get an automatic codename (`Atlas`, `Echo`, `Nova`…) that's unique to the workbench and used as their identity in the UI and in peer messages.
## Shared workspace
```
/workspace/
├── / ← per-agent scratch
├── /
└── shared/ ← anything other agents should see
```
Agents are told about the layout in their system prompt: drop drafts in your own folder, drop hand-offs in `shared/`. References attached at workbench level are materialised into `/workspace/.refs//` and read-only.
## Peer messaging
Agents in the same workbench can message each other with the `peer.send_message` tool. A message from Atlas lands in the recipient's chat tagged with the sender's codename, so the receiver knows it's a hand-off, not a user instruction.
Peer messaging is workbench-only. Standalone agents have no peers.
## Connections and references
Connections and references live on the workbench, not the individual agent. Every agent inherits the set on its first turn, and edits propagate to all active agents on their next turn. Configure the project once instead of per agent.
## Workbench vs single agent
| | Single agent | Workbench |
| ----------- | ------------ | ------------------------------------------ |
| Sandbox | Per task | Shared |
| Filesystem | Per task | Shared (`/workspace//` + `shared/`) |
| Connections | Per task | On the workbench |
| References | Per task | On the workbench |
| Peers | None | Other agents in the workbench |
## Next Steps
The practical guide
The unit running inside a workbench
Shared inputs for every agent
# Adding Integrations
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
Beyond Salesforce, MeshMesh supports several built-in integrations and lets you connect custom MCP servers for additional capabilities.
## Built-in Integrations
Open the **Integrations** menu to see available connections:
**Salesforce Marketing Cloud Engagement** — Connect to MCE for email campaigns, journeys, data extensions, and automations. Requires a Client ID and Secret from an Installed Package with Server-to-Server authentication.
**B2C Commerce Cloud** — Connect to SFCC for code deployment, WebDAV file management, job execution, sandbox management, and more. Requires an Account Manager API Client with OCAPI and WebDAV permissions. See the [B2C Commerce Cloud setup guide](/guides/connect-b2c-commerce) for step-by-step instructions.
**Mulesoft Anypoint Platform** — Connect to Mulesoft for API-led connectivity and integration patterns. Requires a Connected App Client ID and Secret from Anypoint Access Management.
**Image Generation (Gemini)** — Generate images using Google Gemini. Requires a Gemini API key.
**Slack** — Bring your own Slack apps and bots to MeshMesh. Send messages, manage channels, and run workflows. Requires a Bot User OAuth Token (`xoxb-...`) from your Slack App.
**GitHub** — Connect to GitHub for repository access and git operations. Requires a Personal Access Token (PAT). We recommend using a **fine-grained PAT** scoped to only the repositories and permissions your agent needs, with a **short expiration** (e.g. 7 days) for maximum security. You can create one at [GitHub → Settings → Developer settings → Fine-grained tokens](https://github.com/settings/tokens?type=beta).
Always prefer fine-grained PATs over classic tokens. Limit the token to specific repositories and grant only the permissions required (typically **Contents: Read and write**). Set the shortest practical expiration — you can always generate a new token when it expires.
Each integration has its own setup flow. Select it from the Integrations menu and follow the prompts to enter your credentials.
## Custom MCP Servers
MeshMesh supports the Model Context Protocol (MCP), so you can connect any MCP-compatible server to extend your agent's capabilities.
This connects an outside MCP server **into** MeshMesh. To go the other way — drive MeshMesh agents **from** Slack, Claude, or another client — see the [MeshMesh MCP Server](/mcp).
To add a custom MCP, go to **Integrations → Add Custom MCP** and paste your MCP configuration. The format is the same as Claude Desktop and other MCP clients:
```json
{
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-github"],
"env": {
"GITHUB_TOKEN": "ghp_your_token_here"
}
}
```
MeshMesh automatically detects the required credentials and prompts you to fill them in securely.
### Examples
**Jira / Atlassian:**
```json
{
"url": "https://mcp.atlassian.com/v1/mcp",
"type": "http"
}
```
**Braze:**
```json
{
"command": "uvx",
"args": ["--native-tls", "braze-mcp-server@latest"],
"env": {
"BRAZE_API_KEY": "your-braze-api-key",
"BRAZE_BASE_URL": "your-braze-endpoint-url"
}
}
```
Once connected, your agent can use the MCP's tools automatically based on your requests.
Set up your Salesforce connection
Start building with your integrations
# Creating Brand Kits
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
Brand kits capture your brand's voice, tone, style, and templates. When you attach a brand kit to an agent, generated content follows your guidelines.
## Creating a Brand Kit
From the dashboard, scroll to the bottom and click **Manage Brand Kits**. Click **New** to create a brand kit.
Enter your website URL and logo URLs. An agent will automatically analyze your brand and create a brand kit for you — extracting voice, tone, colors, style guidelines, and capturing email and website templates from your existing content.
## Using Brand Kits
When creating an agent, select a brand kit to apply. The agent uses it to keep all generated content on-brand and can reference your templates when creating new emails or pages.
## Next Steps
Help agents understand your organization
Learn about all reference types
# Adding Business Context
import { Card, Cards } from 'fumadocs-ui/components/card';
Business context is background knowledge about your organization. This feature is currently in early preview. It helps agents understand your processes, use your terminology, and work within your existing structures.
## Creating Business Context
From the dashboard, scroll to the bottom and click **Manage Business Context**. Click **New** to create business context.
Select the connections you want to include. An agent will analyze your connected systems and automatically create a business context reference for you — extracting terminology, naming conventions, processes, and organizational knowledge from your existing data.
## Using Business Context
Attach business context to an agent like any other reference. The agent uses it as background knowledge when working on your goals.
Just like any other reference, you can ask your agent to make changes to your business context as your organization evolves.
## Next Steps
Define your brand voice and style
Learn about all reference types
# Chat
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
import { Tabs, Tab } from 'fumadocs-ui/components/tabs';
# Chat with Your AI Assistant
The chat interface is your primary way to interact with MeshMesh. Type what you want to do in natural language, and the AI assistant helps you accomplish your Salesforce tasks.
The chat works like having a conversation with an expert who knows your Salesforce environment
## Conversation Flow
>MeshMesh: Ask question in plain English
MeshMesh->>Salesforce: Query data/Execute action
Salesforce-->>MeshMesh: Return results
MeshMesh-->>User: Present insights & outputs
User->>MeshMesh: Follow-up question
MeshMesh->>Salesforce: Refined query with context
Salesforce-->>MeshMesh: Updated results
MeshMesh-->>User: Contextual response
`}
/>
## How It Works
Query your data in plain English without complex syntax or technical jargon.
**Example:**
```
"Show me all contacts who opened an email in the last 7 days"
```
Receive actionable results with exportable tables, copy-to-clipboard code, and downloadable reports.
**Features:**
* Interactive data tables
* One-click exports
* Runnable code snippets
The AI remembers your conversation and builds on previous requests for complex workflows.
**Example:**
```
User: "Show me VIP customers"
AI: [Shows data]
User: "Now create an email campaign for them"
```
Upload and analyze different types of content including text, data, images, and documents.
**Supported:**
* CSV and Excel files
* Images and PDFs
* Code and configuration files
## Example Conversations
```
User: How many subscribers in each list?
AI: Results:
Newsletter: 12,453
VIP Customers: 2,891
```
```
User: Create a welcome email with first name personalization
AI: I'll create a personalized email:
Hi %%FirstName%%,
Welcome to our community!
```
```
User: Compare email performance for Q3 vs Q4
AI: I'll analyze your email metrics...
Q3 Open Rate: 24.5%
Q4 Open Rate: 28.3%
Improvement: +3.8%
```
## Tips and Tricks
**Be Specific:** Include details like date ranges and data fields for better results
**Use Follow-ups:** Build on previous queries - the AI remembers your conversation context
**Ask for Examples:** Not sure how to phrase something? Ask "show me how to..." for step-by-step guidance
**Export Results:** Save data tables, code snippets, and reports for use in other tools
Learn how to connect your Salesforce products
Master effective AI communication
# Connect B2C Commerce Cloud
import { Callout } from 'fumadocs-ui/components/callout';
import { Tabs, Tab } from 'fumadocs-ui/components/tabs';
import { Accordion, Accordions } from 'fumadocs-ui/components/accordion';
import { Steps, Step } from 'fumadocs-ui/components/steps';
# Connect B2C Commerce Cloud
This guide walks you through connecting your Salesforce B2C Commerce Cloud (SFCC) instance to MeshMesh. Once connected, MeshMesh can deploy code, manage WebDAV files, run jobs, manage sandboxes, and more.
You'll need access to two Salesforce systems:
* **Account Manager** — to create an API Client (ask your admin if you don't have access)
* **Business Manager** — to configure OCAPI and WebDAV permissions
## Prerequisites
Before you begin, make sure you have:
* A B2C Commerce Cloud instance (sandbox or production)
* Access to [Account Manager](https://account.demandware.com) (or an admin who can create an API Client for you)
* Access to Business Manager on your instance
## Step 1: Create an API Client
If you can't access Account Manager, ask your admin to create an API Client for you and send you the **Client ID** and **Client Secret**. They should follow the steps below.
### Log in to Account Manager
Go to [account.demandware.com](https://account.demandware.com) and log in.
### Create the API Client
Navigate to **API Client** in the left menu and click **Add API Client**.
Fill in:
* **Display Name**: A descriptive name (e.g., "MeshMesh Agent")
* **Password**: Generate a strong secret — this is your **Client Secret**. Save it securely.
* **Token Endpoint Auth Method**: Select **`client_secret_basic`**
The B2C CLI only supports `client_secret_basic`. Do not use `client_secret_post` or `private_key_jwt`.
### Configure Default Scopes
Under **Default Scopes**, add these scopes:
| Scope | Purpose |
| -------------- | ------------------------------------- |
| `mail` | Required for authentication flows |
| `roles` | Returns role information in the token |
| `tenantFilter` | Returns tenant access information |
| `openid` | Required for OpenID Connect |
Without `roles` and `tenantFilter`, the OAuth token comes back empty and nothing works.
### Assign Roles
Under the API Client's **Roles** section, add:
| Role | When Needed | Tenant Filter |
| --------------------------- | ---------------------------------------------- | ---------------------------------- |
| **Salesforce Commerce API** | SCAPI commands, eCDN, CIP analytics | Your tenant IDs (e.g., `zzxy_prd`) |
| **Sandbox API User** | ODS sandbox management, SLAS client management | Your realm/org IDs |
For each role, configure the **Tenant Filter** with the tenant IDs or realm IDs you need to access.
## Step 2: Configure OCAPI Permissions
OCAPI permissions are needed for code management, job execution, and site operations. These are configured per-instance in Business Manager.
### Open OCAPI Settings
In Business Manager, navigate to **Administration > Site Development > Open Commerce API Settings** and select the **Data API** type.
### Add the Configuration
Add a JSON configuration for your API Client ID:
```json
{
"_v": "24.5",
"clients": [
{
"client_id": "your-client-id",
"resources": [
{
"resource_id": "/code_versions",
"methods": ["get"],
"read_attributes": "(**)",
"write_attributes": "(**)"
},
{
"resource_id": "/code_versions/*",
"methods": ["get", "put", "patch", "delete"],
"read_attributes": "(**)",
"write_attributes": "(**)"
},
{
"resource_id": "/jobs/*/executions",
"methods": ["post"],
"read_attributes": "(**)",
"write_attributes": "(**)"
},
{
"resource_id": "/jobs/*/executions/*",
"methods": ["get"],
"read_attributes": "(**)",
"write_attributes": "(**)"
},
{
"resource_id": "/job_execution_search",
"methods": ["post"],
"read_attributes": "(**)",
"write_attributes": "(**)"
},
{
"resource_id": "/sites",
"methods": ["get"],
"read_attributes": "(**)",
"write_attributes": "(**)"
},
{
"resource_id": "/sites/*",
"methods": ["get"],
"read_attributes": "(**)",
"write_attributes": "(**)"
}
]
}
]
}
```
Replace `your-client-id` with your actual API Client ID.
```json
{
"resource_id": "/code_versions",
"methods": ["get"],
"read_attributes": "(**)",
"write_attributes": "(**)"
},
{
"resource_id": "/code_versions/*",
"methods": ["get", "put", "patch", "delete"],
"read_attributes": "(**)",
"write_attributes": "(**)"
}
```
```json
{
"resource_id": "/jobs/*/executions",
"methods": ["post"],
"read_attributes": "(**)",
"write_attributes": "(**)"
},
{
"resource_id": "/jobs/*/executions/*",
"methods": ["get"],
"read_attributes": "(**)",
"write_attributes": "(**)"
},
{
"resource_id": "/job_execution_search",
"methods": ["post"],
"read_attributes": "(**)",
"write_attributes": "(**)"
}
```
## Step 3: Configure WebDAV Client Permissions
WebDAV permissions are needed for file operations like code deployment, site import/export, and log access.
### Open WebDAV Settings
In Business Manager, navigate to **Administration > Organization > WebDAV Client Permissions**.
### Add the Configuration
Add this JSON configuration for your API Client ID:
```json
{
"clients": [
{
"client_id": "your-client-id",
"permissions": [
{"path": "/cartridges", "operations": ["read_write"]},
{"path": "/impex", "operations": ["read_write"]},
{"path": "/logs", "operations": ["read_write"]}
]
}
]
}
```
Replace `your-client-id` with your actual API Client ID.
| Path | Used For |
| ------------- | ------------------ |
| `/cartridges` | Code deployment |
| `/impex` | Site import/export |
| `/logs` | Log file access |
Add more paths as needed (e.g., `/catalogs/` for catalog files, `/libraries/` for content libraries).
## Step 4: Add the Integration in MeshMesh
### Open Integrations
In MeshMesh, go to **Integrations** and select **B2C Commerce Cloud**.
### Enter Your Credentials
Fill in the required fields:
| Field | Value |
| ------------------- | ---------------------------------------------------------- |
| **Server Hostname** | Your instance hostname (e.g., `myinstance.demandware.net`) |
| **Client ID** | The API Client ID from Account Manager |
| **Client Secret** | The password you set on the API Client |
Optional fields for advanced operations:
| Field | When Needed |
| ---------------- | -------------------------------------------- |
| **Code Version** | Default code version for deployments |
| **Tenant ID** | For SCAPI/eCDN commands (format: `zzxy_prd`) |
| **Short Code** | For SCAPI commands |
| **MRT API Key** | For Managed Runtime deployments |
### Attach to a Task
Add the B2C Commerce integration to any task that needs to interact with your Commerce Cloud instance.
## What MeshMesh Handles Automatically
Once the integration is added and attached to a task:
* **OAuth authentication** is handled per command — no manual login needed
* **Server hostname** is auto-injected — no need to specify `--server`
* **Credentials are secure** — the agent never sees your Client Secret
* **Token refresh** happens automatically on each interaction
Just describe what you want to do in natural language, and MeshMesh takes care of the rest.
## Troubleshooting
* Verify your Client ID and Client Secret are correct
* Check that OCAPI is configured for your Client ID in Business Manager
* Ensure the API Client has the required roles in Account Manager
* Confirm the Token Endpoint Auth Method is `client_secret_basic`
* Check WebDAV Client Permissions in Business Manager
* Ensure the paths you're accessing are listed in the permissions
* Verify the Client ID matches in both OCAPI and WebDAV configurations
* Add `mail`, `roles`, `tenantFilter`, and `openid` to Default Scopes
* For SCAPI commands, add the relevant `sfcc.*` scopes
* Do NOT add `SALESFORCE_COMMERCE_API` as a scope (it's a role, not a scope)
* Change the Token Endpoint Auth Method to `client_secret_basic` in Account Manager
* The B2C CLI does not support `client_secret_post` or `private_key_jwt`
## Optional: SCAPI Scopes
For advanced operations (eCDN, SCAPI schemas, custom APIs), add these scopes to your API Client's Default Scopes:
| Scope | Purpose |
| -------------------- | --------------------- |
| `sfcc.cdn-zones` | eCDN read operations |
| `sfcc.cdn-zones.rw` | eCDN write operations |
| `sfcc.scapi-schemas` | SCAPI schema browsing |
| `sfcc.custom-apis` | Custom API status |
# Connecting Marketing Cloud Engagement
import { Callout } from 'fumadocs-ui/components/callout';
import { Steps, Step } from 'fumadocs-ui/components/steps';
# Connecting Marketing Cloud Engagement
This guide walks you through connecting your Salesforce Marketing Cloud Engagement (SFMC) account to MeshMesh. Once connected, MeshMesh can manage Data Extensions, Journeys, Automations, Email content, Transactional Sends, and more.
MeshMesh uses the **Public App** integration type with **Authorization Code** grant, so no client secret is needed. Each user authenticates with their own SFMC account, providing a full audit trail.
* Admin access to your Marketing Cloud Engagement account (to create an Installed Package)
* Your Marketing Cloud subdomain (the `mc...` part of your API URLs)
## Prerequisites
* A Salesforce Marketing Cloud Engagement account
* Admin access to **Setup > Platform Tools > Apps > Installed Packages**
* Your tenant-specific subdomain (found in your MC API URLs)
## Step 1: Create an Installed Package
Log in to Marketing Cloud Engagement and navigate to **Setup > Platform Tools > Apps > Installed Packages**.
Click **New** to create a new Installed Package.
Give it a name (e.g., "MeshMesh Studio") and an optional description. Click **Save**.
## Step 2: Add a Public App API Integration
In your new Installed Package, click **Add Component**.
Select **API Integration** and click **Next**.
Select **Public App** as the integration type. This is important — Public App does not require a client secret, which is more secure for desktop applications.
Configure the integration:
* **Redirect URI**: Pick the one that matches where your team uses MeshMesh. Marketing Cloud only allows a single redirect URI per Installed Package.
* Desktop app: `https://auth.meshmesh.io/sfmc-callback`
* Web app: `https://studio.meshmesh.io/auth/sfmc-callback`
If your team uses both, you'll need a separate Installed Package (and a separate MeshMesh connection) for each — see the note below.
* **Scope**: Select the scopes your team needs. For full MeshMesh functionality, enable:
* `Data Extensions` (Read, Write)
* `Automations` (Read, Write, Execute)
* `Journeys` (Read, Write, Execute)
* `Email` (Read, Write, Send)
* `Contacts` (Read, Write)
* `Tracking Events` (Read)
* `Saved Content` (Read, Write)
* `List and Subscribers` (Read, Write)
Click **Save**. Note the **Client ID** — you'll need it in MeshMesh.
Since you selected **Public App**, there is no client secret. This is the correct and most secure configuration for MeshMesh. Do not use "Web App" or "Server-to-Server" integration types.
## Step 3: Find Your API Base URIs
After saving, your Installed Package summary page will show:
* **Authentication Base URI**: e.g., `https://mc1234567890.auth.marketingcloudapis.com`
* **REST Base URI**: e.g., `https://mc1234567890.rest.marketingcloudapis.com`
Copy both of these URIs.
## Step 4: Connect in MeshMesh
In MeshMesh Studio, click **Integrations** in the top bar.
Select **Salesforce Marketing Cloud Engagement** from the list.
Enter a name for the connection (e.g., "Production MCE").
Fill in the Installed Package details:
* **Client ID**: From your Installed Package
* **Auth Base URI**: Your authentication base URI
* **REST Base URI**: Your REST API base URI
* **Account ID (MID)**: Optional — only needed if using multiple business units
Click **Connect with Marketing Cloud**. Your browser will open to the Marketing Cloud login page.
Log in with your Marketing Cloud credentials. After successful authentication, you'll be redirected back to MeshMesh Studio.
Each user who needs to use the SFMC integration must authenticate with their own Marketing Cloud account. This ensures all actions are traceable to individual users.
## How Authentication Works
MeshMesh uses the OAuth 2.0 **Authorization Code** grant type with a **Public App** integration:
1. You click "Connect" and log in to Marketing Cloud via your browser
2. Marketing Cloud sends an authorization code to MeshMesh's callback server
3. The server exchanges the code for an access token and refresh token
4. The refresh token is stored securely in MeshMesh's encrypted vault
5. Only short-lived access tokens (\~20 minutes) are sent to the agent sandbox
6. Tokens are automatically refreshed — no manual re-authentication needed
**Security highlights:**
* No client secret is used or stored anywhere
* Refresh tokens are stored server-side only, never on your device
* Access tokens expire in \~20 minutes and are refreshed automatically
* All actions use your personal Marketing Cloud identity for full audit trails
* Refresh token rotation is supported — if Marketing Cloud issues a new refresh token, it's automatically updated
## Troubleshooting
### "Invalid redirect URI" error
Your Installed Package's redirect URI must match the MeshMesh app you're connecting from:
* `https://auth.meshmesh.io/sfmc-callback` for the desktop app
* `https://studio.meshmesh.io/auth/sfmc-callback` for the web app
Marketing Cloud only allows one redirect URI per Installed Package. If you originally set up your Installed Package for the desktop app and now want to connect from the web app (or vice versa), you'll need to create a **new Installed Package** with the matching redirect URI and a new MeshMesh connection that uses it. Your existing connection will keep working on the surface it was set up for.
### Using MeshMesh from both desktop and web
Because each Installed Package can register only one redirect URI, set up two Installed Packages — one with the desktop URI, one with the web URI — and create a separate MeshMesh connection for each (e.g. "Production MCE — Desktop" and "Production MCE — Web"). Both can authenticate the same Marketing Cloud users.
### "Insufficient permissions" error
Check that your Installed Package has the required scopes enabled, and that your Marketing Cloud user account has the necessary permissions.
### Connection expires
Access tokens are refreshed automatically. If your refresh token expires (e.g., after extended inactivity), you'll need to reconnect by clicking "Connect" again in the Integrations panel.
### Multi-Business Unit setup
If your Marketing Cloud account has multiple business units, enter the Account ID (MID) of the business unit you want to connect to. You can create separate connections for different business units.
# Connecting a Salesforce Org
import { Callout } from 'fumadocs-ui/components/callout';
import { Steps } from 'fumadocs-ui/components/steps';
MeshMesh connects to Salesforce using OAuth. You authenticate through your browser — no passwords are stored. There are several ways to connect, depending on your setup:
* **Quick Connect** — fastest, desktop only, no admin setup.
* **External Client App** — your own app, works **headless** (web app, Slack, hosted MCP clients) as well as desktop.
* **Custom Connected App** — your own app credentials via the Salesforce CLI, desktop only.
* **MeshMesh Connected App** — a managed package an admin installs once (pre-release).
## Quick Connect (Preferred)
**The fastest way to connect — just click and go.**
No admin setup or package installation required. Your Salesforce user just needs **API permissions** (enabled by default for most profiles).
### Connect from MeshMesh Studio
Open MeshMesh Studio desktop → **Connections** → select **Quick Connect** → click **Connect**.
### Authorize in browser
Sign in through the browser and approve access. Done.
Quick Connect uses the Salesforce CLI and is desktop only. To use MeshMesh **on the web** or from **Slack**, use the External Client App method below.
## External Client App
**Use your own Salesforce External Client App — the only method that works headless.**
This is the method to choose if you want to use MeshMesh **without the desktop app**: on the web, from the **Slack** integration, or with other hosted MCP clients. You create an External Client App (ECA) in your own org and register a MeshMesh callback URL on it. Because MeshMesh stores and refreshes the OAuth token **server-side** (no Salesforce CLI, no laptop, no desktop app required), the connection keeps working in fully server-side / hosted contexts.
### Create an External Client App in Salesforce
In Salesforce Setup, go to **External Client App Manager** → **New External Client App**. Then:
* **Enable OAuth Settings** (use the defaults for the other fields).
* **OAuth Scopes** — add at minimum:
* `api`
* `refresh_token`
* `offline_access`
* `web`
Add any other scopes your use case needs.
### Configure security settings
* **Require Proof Key for Code Exchange (PKCE)**: leave **enabled**.
* **Enable Client Credentials Flow**: leave this **off** — MeshMesh uses the authorization-code flow, not client credentials.
* A consumer secret is **not** required — MeshMesh connects as a public PKCE client.
### Set the Callback URL
Set the app's **Callback URL** to the exact URL shown in the MeshMesh Connections modal for your platform and environment:
| Where you run MeshMesh | Callback URL |
| ---------------------- | ----------------------------------------------------- |
| Web (production) | `https://studio.meshmesh.io/auth/sf-callback` |
| Web (staging) | `https://studio-staging.meshmesh.io/auth/sf-callback` |
| Desktop (production) | `https://auth.meshmesh.io/sf-callback` |
| Desktop (staging) | `https://staging-auth.meshmesh.io/sf-callback` |
If you use MeshMesh on both web and desktop, register **both** callback URLs on the app. The Connections modal always shows the exact value(s) to copy for your current context.
### Wait for activation
A new External Client App can take **up to 30 minutes** to become available. If the connection fails right after creation, wait and try again.
### Copy the Consumer Key
In the app's **Settings → OAuth Settings**, copy the **Consumer Key**. This is a public identifier, not a secret.
### Connect from MeshMesh
Open **Connections** → select **External Client App** → paste your **Consumer Key** → pick the environment → click **Connect** and authorize in the browser.
Pick this method for the web app, Slack, or any hosted client. MeshMesh uses whatever OAuth scopes you configured on your External Client App, so the app itself is the source of truth for access.
## Custom Connected App
**For orgs that require their own OAuth credentials or specific scopes, on desktop.**
Use this method on the **desktop app** if your org has restrictions on third-party apps or you need fine-grained control over OAuth scopes. It authenticates through the Salesforce CLI, so it does not work on the web or from Slack — use the **External Client App** method above for those.
### Create a Connected App in Salesforce
In Salesforce Setup, go to **App Manager** → **New Connected App**. Configure the following:
* **Enable OAuth Settings**: Check this box
* **Callback URL**: `http://localhost:1717/OauthRedirect` (the Salesforce CLI loopback callback)
* **OAuth Scopes**: Add at minimum:
* `api`
* `refresh_token`
* `offline_access`
* `web`
Add `full` or other scopes as needed for your use case.
### Copy credentials
Save the **Consumer Key** (Client ID) — and the **Consumer Secret** if your app requires one — from the connected app settings.
### Enter credentials in MeshMesh Studio
Open MeshMesh Studio desktop → **Connections** → select **Custom Connected App** → expand **OAuth Scopes & Credentials** → enter your Client ID (and Client Secret if required).
### Connect and authorize
Click **Connect**, sign in through the browser, and approve access.
Use this method if your org blocks third-party connected apps or if you need fine-grained control over which OAuth scopes are granted, and you connect from the desktop app.
## MeshMesh Connected App (Pre-release, Supported Clients Only)
This method is currently in **pre-release** and only available to supported clients. It may change in future releases.
**Best for teams — no setup needed.**
The MeshMesh Connected App method uses a managed package that installs an External Client App (ECA) in your Salesforce org. An admin installs it once, then everyone in the org can connect.
### Install the MeshMesh Studio package
A Salesforce admin installs the MeshMesh Studio package in the target org. Choose "Install for All Users" when prompted.
Use the URL that matches **the org you're installing into** — only the host differs, the package id is the same everywhere:
* **Production:** [`https://login.salesforce.com/packaging/installPackage.apexp?p0=04tfj000000GGiLAAW`](https://login.salesforce.com/packaging/installPackage.apexp?p0=04tfj000000GGiLAAW)
* **Sandbox:** [`https://test.salesforce.com/packaging/installPackage.apexp?p0=04tfj000000GGiLAAW`](https://test.salesforce.com/packaging/installPackage.apexp?p0=04tfj000000GGiLAAW)
* **Custom My Domain:** `https:///packaging/installPackage.apexp?p0=04tfj000000GGiLAAW` — substitute your org's My Domain host. Example: `https://acme.my.salesforce.com/packaging/installPackage.apexp?p0=04tfj000000GGiLAAW`. Sandbox my-domains follow the same pattern, e.g. `https://acme--uat.sandbox.my.salesforce.com/packaging/installPackage.apexp?p0=04tfj000000GGiLAAW`.
Log into the target org in your browser **before** opening the install URL. If you're signed into a different org, the package lands there instead.
### Enable the External Client App
In Salesforce Setup, go to **External Client App Manager**, find **MeshMesh Studio**, and click **Enable**.
### Connect from MeshMesh Studio
Open MeshMesh Studio desktop → **Connections** → select **MeshMesh App** as the connection method → click **Connect**.
### Authorize in Salesforce
Sign in through the browser and approve access. You'll be redirected back to MeshMesh Studio.
The package install and enable steps are one-time admin setup. After that, any user in the org can connect without additional configuration.
## Sandbox & Custom Domains
MeshMesh supports production orgs, sandbox orgs, and custom domains (My Domain). Two things have to match the org you're connecting to: the **package install URL** (see the install step above) and the **environment picker** in the Connections modal.
**For sandbox orgs:** Select **Sandbox** in the environment picker before connecting. This directs authentication to `test.salesforce.com`. Make sure you installed the package via the sandbox install URL — `https://test.salesforce.com/packaging/installPackage.apexp?p0=04tfj000000GGiLAAW` — not the production one.
**For custom domains:** Select **Custom Domain** and enter your instance URL (e.g., `https://MyDomain--SandboxName.sandbox.my.salesforce.com`). The package install URL needs to use that same host:
```
https:///packaging/installPackage.apexp?p0=04tfj000000GGiLAAW
```
If you install the package in production but pick **Sandbox** (or vice versa) when connecting, OAuth fails with "grant type not supported" — the auth code was issued by one host and the token endpoint can't redeem it on another.
## Troubleshooting
**"External client app not installed"** — Install the MeshMesh Studio package in your org and enable it in **Setup → External Client App Manager**. Use the install URL matching your environment:
* Production: [`https://login.salesforce.com/packaging/installPackage.apexp?p0=04tfj000000GGiLAAW`](https://login.salesforce.com/packaging/installPackage.apexp?p0=04tfj000000GGiLAAW)
* Sandbox: [`https://test.salesforce.com/packaging/installPackage.apexp?p0=04tfj000000GGiLAAW`](https://test.salesforce.com/packaging/installPackage.apexp?p0=04tfj000000GGiLAAW)
* Custom My Domain: `https:///packaging/installPackage.apexp?p0=04tfj000000GGiLAAW`
**"Grant type not supported"** — The package was installed in a different org than the one you're trying to connect to (e.g. installed in production but connecting to a sandbox). Either install the package in the right org using the URL matching your environment above, or change the **Environment** selector in the Connections modal to match where the package is installed.
**"Cross-org OAuth error"** — The MeshMesh Studio package must be both installed *and* enabled in the org you're connecting to. Go to **Setup → External Client App Manager → MeshMesh Studio → Enable**.
**"Authorization failed"** — Verify that your External Client App or Connected App has the required OAuth scopes (`api`, `refresh_token`, `offline_access`, `web`).
**"Salesforce CLI required"** — Quick Connect and Custom Connected App both use the Salesforce CLI (desktop only). To connect without the CLI — including on the web or from Slack — use the **External Client App** method instead.
**"redirect\_uri mismatch" / "invalid redirect URI"** (External Client App) — The Callback URL registered on your External Client App must exactly match the one shown in the Connections modal for your platform and environment (web vs desktop, production vs staging). Copy the value from the modal and register it verbatim; if you use both web and desktop, register both.
**Connection timeout** — Check your network and firewall settings. Ensure your browser can reach `login.salesforce.com` (or `test.salesforce.com` for sandboxes).
# Creating Skills
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
Beyond the built-in [skills](/guides/using-skills) that ship with every agent, you can add your own. A skill is a capability pack — a `SKILL.md` file plus any supporting scripts and assets — that lives in your [Library](/concepts/library) as a reference.
There are two ways to add one: **upload** a skill you've built, or **import** skills from a git repo.
## Upload a skill
1. In the Library, choose **New → Reference**, then **Upload Files**.
2. Drag in a folder whose root contains a `SKILL.md` (alongside any scripts or assets it needs).
3. MeshMesh detects the `SKILL.md` and classifies the reference as a **Skill** automatically. Give it a name and create it.
The `SKILL.md` must be at the **root** of the folder you upload. If it's nested in a subfolder, MeshMesh warns you and won't treat the upload as a skill.
## Import skills from a git repo
Workspace **admins** and **owners** can pull skills straight from a public git repository.
1. In **Discover**, open the **Skills** collection.
2. Click **Add skills from repo**.
3. Point it at a repository. MeshMesh discovers the `SKILL.md` files it contains.
4. Select the skills to import.
Imported skills go into the **Review** queue and become available once published.
Import currently supports **public** repositories.
## How skills are governed
Every uploaded or imported skill is checked before it can be used workspace-wide:
* **Scanned** for safety when it's added.
* **Reviewed** — imports enter the admin publish-review queue before they go live.
* **Version-pinned** — imported skills are pinned to the exact commit they were discovered at, so what runs is always what was reviewed.
* **Scoped** to your workspace and to whoever it's shared with.
If a `SKILL.md` tries to grant itself tool permissions, those are stripped when the skill is delivered to an agent — a skill can't widen what an agent is allowed to do.
## Next Steps
How agents pick and use skills
Put a skill to work in an agent
# Creating Your First Agent
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
In MeshMesh, an **agent** is an AI assistant designed to accomplish a specific goal or objective. While you can use MeshMesh as a chatbot for quick questions, it's really built to execute work for you — implementing a Sales Cloud configuration, building an Agentforce assistant, analyzing your pipeline, or cleaning up data across your org.
When you create an agent, you're giving it a mission. The agent figures out the steps needed, connects to your Salesforce org, and works through the objective. You can watch its progress, provide guidance when it asks questions, and review what it produces.
## Start from a Blueprint
Blueprints are pre-built templates that give your agent a head start. Each blueprint is optimized for a specific type of work and includes context the agent uses to produce better results.
Configure leads, accounts, contacts, opportunities, sales processes, and pipeline dashboards
Get insights on pipeline coverage, deal velocity, at-risk opportunities, and forecast accuracy
Review custom code, object model, automation, and security configurations
To use a blueprint, select it when creating a new agent. The blueprint provides a starting point — you'll customize it with your specific requirements.
## Start from Scratch
When blueprints don't fit your needs, create an agent from scratch by describing what you want in your own words. This gives you full flexibility for custom requirements, cross-product workflows, or exploratory work where you're not sure exactly what you need yet.
A good description has three parts: the goal (what you want to accomplish), the context (relevant details and constraints), and the expected output (what you want to receive).
Here's an example:
```
Identify customers at risk of churning.
Use Data Cloud customer records. Define "at risk" as no purchase
in 90 days but still engaging with emails. Exclude anyone with
an open support ticket.
Return a list with customer name, last purchase date, and
engagement score, sorted by highest engagement first.
```
The goal is clear, the context provides necessary details, and the expected output is specific.
## Plan Mode vs Build Mode
MeshMesh offers two ways to work depending on how much oversight you want.
**Plan Mode** gives the agent read-only access to your connections. It can analyze your org, access the web for research, and create a detailed plan — but it won't make any changes until you switch to Build Mode. You'll see exactly what it intends to do, review the plan, ask for modifications, and then decide whether to proceed.
**Build Mode** lets the agent make changes to your org if you ask it to. The agent starts executing as soon as you describe what you want — faster for straightforward tasks when you already know what you need.
You can switch between modes anytime from the mode selector.
# Working with Multiple Orgs
import { Callout } from 'fumadocs-ui/components/callout';
import { Steps } from 'fumadocs-ui/components/steps';
When working across multiple Salesforce orgs — whether you're migrating configurations, duplicating setups, or using one org as a reference for building another — it's important to follow a workflow that keeps your orgs safe and separate.
**Cross-org pollination risk.** If the same agent has write access to both a reference org and a target org, there's a possibility of unintended changes being made to your reference org. This is especially important for larger integrations where the scope of changes is significant.
## Best Practice: Two-Agent Workflow
The recommended approach for migrating or duplicating items from a read-only reference org to a new target org is to use **two separate agents** — one for planning, one for building.
### Assess with your reference org in Plan Mode
Start a new agent with your **reference org** connected. Set the agent to **Plan Mode** so it has read-only access. Have the agent assess what you want to migrate and capture everything into a detailed plan.
This gives you a complete picture of the source org's configuration without any risk of modifying it.
### Build in your target org using the plan
Start a **second agent** with your **target org** connected. Select the existing plan from the first agent as a reference, then run your build.
The second agent executes against only the target org, using the plan as its guide. Your reference org stays untouched.
## Why This Works
This two-agent approach provides clear separation between reading and writing:
* **The reference org is never exposed to write operations.** Plan Mode ensures the first agent can only read and analyze — it cannot make changes.
* **The target org gets a focused build.** The second agent works from a well-defined plan, reducing the chance of errors or scope creep.
* **No shared agent context between orgs.** Each agent operates independently with its own connection, eliminating any possibility of cross-org actions.
**Tips for a smooth workflow:**
* Review and refine the plan from the first agent before starting the second agent. A thorough plan leads to a cleaner build.
* Keep the reference agent open while building so you can quickly check details or ask follow-up questions about the source org.
* For complex migrations, break the work into smaller plans that each cover a specific area (e.g., object model, automations, permissions) rather than one massive migration plan.
# Creating and Sharing Recipes
import { Callout } from 'fumadocs-ui/components/callout';
import { Steps, Step } from 'fumadocs-ui/components/steps';
import { Card, Cards } from 'fumadocs-ui/components/card';
Recipes let you capture work you've already done with an agent and turn it into a reusable guide. Once saved, you can share recipes with your team or attach them to future agents to repeat the same workflow.
## Creating a Recipe
### Complete your work
First, work with an agent to complete a task — set up an integration, configure a feature, build a report, or anything else you want to make repeatable.
### Ask the agent to capture it
Once you're happy with the result, tell the agent to turn it into a recipe. You can say things like:
* *"Turn this into a recipe"*
* *"Save this as a runbook"*
* *"Make this an SOP"*
* *"Document what we just did"*
### Answer the scoping questions
The agent will ask you two things:
* **What to include** — Should the recipe cover everything, or just specific parts of the work?
* **Exact vs. generalized** — Should it keep your specific details (for identical reimplementation), or use placeholders so it works across different projects?
### Recipe is saved
The agent reviews the full conversation, distills the key steps, and saves the recipe as a reference. It automatically adds relevant tags and a description so others can discover it.
## Sharing Recipes
Once a recipe is created, you can share it just like any other reference:
* **With specific people** — Open the reference, click Share, and add team members by name or email
* **With your organization** — Set the access to your organization name so everyone on your team can see it
* **Via link** — Copy the share link and send it directly
Shared recipes appear in the **Shared** tab of the References panel. Team members can star recipes they use frequently for quick access.
## Discovering Recipes
Open the **References** panel and use the sidebar to browse:
* Filter by **Recipes** in the type list to see all available recipes
* Switch to the **Shared** tab to see recipes shared by teammates
* Use **tags** to find recipes for specific products or operation types
* **Star** recipes you use often so they appear at the top of your list
## Using a Recipe
To use a recipe, attach it to a new agent as a reference. The agent reads the recipe and follows the steps to reproduce the same outcome. If the recipe uses placeholders, the agent will ask you to provide the specific values for your project.
Learn more about the recipe concept
How to attach references to agents
# Search & Navigation
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
import { Tabs, Tab } from 'fumadocs-ui/components/tabs';
# Search & Navigation
MeshMesh gives you two ways to search: the **CMD+K command palette** for quick access from anywhere, and the **dashboard search bar** for filtered browsing.
## CMD+K Command Palette
Press **`Cmd+K`** (Mac) or **`Ctrl+K`** (Windows/Linux) from anywhere in the app to open the command palette. It searches across all your agents, references, and blueprints in one place.
### What You Can Do
Search by agent name or conversation content. CMD+K searches both local agent names and server-side conversation history, so you can find agents by what was discussed — not just what they were named.
When you select an agent from the results, you'll be taken directly to its conversation. If the match came from a specific message, you'll be scrolled to that exact message.
Search references by name, description, or type. Results show the reference type so you can quickly distinguish between documents, images, and other file types.
Selecting a reference navigates to the dashboard and opens its preview.
Search blueprints by name, description, or category. Results display the blueprint category for easy identification.
Selecting a blueprint navigates to the dashboard and opens its preview.
With no search query, CMD+K shows a **Start New Agent** quick action at the top. You can also type a name and start an agent with that name directly — useful when no existing results match what you're looking for.
**Search conversation content:** Type at least 2 characters to trigger a deep search across all agent conversations — not just agent names.
**Keyboard navigation:** Use arrow keys to move through results and Enter to select.
**Quick dismiss:** Press Escape or click outside to close the palette.
## Dashboard Search
The search bar on the dashboard provides a more detailed search experience with filters. Click the search bar or start typing to expand it.
### Filters
Dashboard search supports filtering by:
* **Yours / Public** — toggle between your own items and publicly shared ones
* **Product** — filter by Salesforce product (Sales Cloud, Service Cloud, etc.)
* **Category** — filter blueprints by category
* **Connection Type** — filter by the type of Salesforce connection required
Filters and search work together — you can search for "email" while filtering to Marketing Cloud blueprints only.
### What Gets Searched
| Item | Searched Fields |
| -------------- | --------------------------- |
| **Agents** | Name, conversation content |
| **References** | Name, description, type |
| **Blueprints** | Name, description, category |
Results appear instantly for local items. Server-side results (public references, blueprints, and conversation content) load with a brief delay.
## Keyboard Shortcuts
| Shortcut | Action |
| --------------- | ------------------------------- |
| `Cmd/Ctrl + K` | Open command palette |
| `Escape` | Close command palette or search |
| `Arrow Up/Down` | Navigate search results |
| `Enter` | Select highlighted result |
Learn how to interact with the chat interface
Get started with your first MeshMesh agent
# Using References with Agents
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
Agents work better when you give them context. References are documents, files, and requirements you attach to help the agent understand what you need and how to deliver it.
You might attach a product brief so the agent knows the features to highlight, a brand guide so it matches your tone, or a spreadsheet with data to analyze. The more relevant context you provide, the better the output.
## Supported Formats
MeshMesh supports common document formats:
* **Documents** — `.docx`, `.pdf`, `.txt`, `.md`
* **Spreadsheets** — `.xlsx`, `.csv`
* **Images** — `.png`, `.jpg`, `.gif`
The agent reads and understands the content, using it to inform its work.
## Adding References
You can add references when creating an agent from scratch, or anytime while you're working in an agent. There are three ways to do it:
**Select existing references** — Choose from documents you've used in other agents. Useful for standard materials like brand guides or process docs that apply across multiple projects.
**Upload new files** — Drag and drop files directly, or click to browse your computer. Uploaded files are saved and available for future agents.
**Create a new document** — Write or paste content directly in MeshMesh Studio. Helpful when you want to draft requirements on the fly or compile notes without leaving the app.
You don't need to attach everything. Focus on materials that directly relate to the agent's objective — requirements, examples, data, or constraints.
## What to Include
**Requirements and specs** — What does the agent need to deliver? Include acceptance criteria, constraints, and must-haves.
**Examples** — Show the agent what good looks like. A sample email, a previous report, or a template to follow.
**Data** — Spreadsheets or CSVs the agent should analyze, reference, or use as input.
**Brand and style guides** — Tone of voice, terminology preferences, formatting rules.
**Process documentation** — How things work in your org, approval flows, naming conventions.
## Best Practices
**Break large requirements into phases** — If you have a complex project with extensive requirements, don't try to do everything at once. Split it into subtasks or phases. Prompt the agent to focus on one phase at a time: "Plan the first phase" or "Build the first phase only." This gives you checkpoints along the way and makes it easier to course-correct.
**Keep images lightweight** — Large, high-resolution images slow things down and can hit size limits. Resize or compress images before uploading. If the agent just needs to understand what's in the image, a smaller version works fine.
**Use accurate data** — The agent trusts what you give it. If you provide sample data, make sure it reflects reality. Placeholder or fake data can lead the agent to make incorrect assumptions or produce outputs that don't match your actual use case.
## Editing and Downloading References
References aren't static — they can evolve as your project progresses.
**Manual editing** — For text-based files like `.md`, `.html`, `.json`, and `.txt`, you can click on a reference to open it, make your changes inline, and save. The agent will use the updated version going forward.
**Agent-assisted editing** — For binary files like `.docx` and `.xlsx`, ask the agent to make changes for you. Just describe what you want updated and the agent will modify the file directly.
You can also ask the agent to create new references from scratch. If you're working through a project and want to capture decisions or document requirements, ask the agent to write them to a reference. This is a great way to build up project documentation as you go.
**Downloading** — Any reference can be downloaded to your computer. Click the download button on the reference to save a local copy.
Think of references as a shared workspace between you and the agent. The more you use them, the more context the agent has to work with.
# Using Agent Skills
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
Skills give the agent deep expertise in specific Salesforce products. When the agent uses a skill, it has access to detailed knowledge about APIs, best practices, and implementation patterns.
## Built-in and custom skills
There are two kinds of skills:
* **Built-in skills** ship with every agent — the Data Cloud, Agentforce, Marketing Cloud, Metadata, and other product skills described below. You don't install them; the agent just uses them.
* **Custom skills** are ones you add to your workspace's [Library](/concepts/library) — upload your own, or (as an admin) import them from a git repo. See [Creating Skills](/guides/creating-skills).
Either way, the agent selects and uses skills the same way.
## How Agent Skills Work
Most of the time, the agent automatically selects the right skill based on your request. If you mention "Data Cloud," the Data Cloud skill activates. If you ask about "deploy my Flow," the Metadata skill kicks in.
You can also explicitly request a skill:
```
Use the Data Cloud skill to set up identity resolution
for my customer profiles
```
```
Use the Agentforce skill to create a service assistant
that can look up order status
```
Skills work with blueprints. If you're following a blueprint, the relevant skills are automatically available to the agent.
## Data Cloud Skill
The Data Cloud skill knows how to work with Salesforce Data Cloud APIs. Use it for:
* **Data streams** — Ingest data from CRM, Marketing Cloud, external sources
* **Connections** — Set up IngestAPI, MCE, CRM, Snowflake, S3 connectors
* **Segments** — Create and publish customer segments
* **Activation targets** — Export data to S3, Azure, GCS, SFTP
* **Identity resolution** — Unify customer profiles across sources
* **Query** — Run SQL queries against your data lake
**Example prompts:**
```
Create a Data Cloud segment for customers who purchased
in the last 90 days but haven't opened an email in 30 days
```
```
Configure identity resolution using email as the primary
match key with fallback to phone number
```
## Agentforce Skill
The Agentforce skill builds AI agents in Salesforce. Use it for:
* **Bots** — Create and deploy conversational agents
* **GenAI Functions** — Define individual actions the agent can take
* **Plugins** — Group functions with instructions
* **Planners** — Orchestrate how the agent decides what to do
* **Apex integration** — Connect functions to Apex invocable methods
**Example prompts:**
```
Create an Agentforce service agent that can look up orders,
check shipping status, and create return requests
```
```
Build a customer support bot that uses our knowledge base
to answer questions
```
The Agentforce skill includes a scaffold tool that generates a complete agent structure. Just describe what you need.
## Marketing Cloud Skill
The Marketing Cloud skill handles all Marketing Cloud Engagement operations. Use it for:
* **Content Builder** — Create and manage emails, images, content blocks
* **Data extensions** — Create, query, insert, update, upsert data
* **Journeys** — Fire entry events, manage journey configurations
* **Automations** — Create and manage automation studio workflows
* **Transactional messaging** — Send real-time triggered emails
* **Subscribers** — Manage subscriber records and lists
**Example prompts:**
```
Create an HTML email in Content Builder with our
standard header and footer
```
```
Create an automation that runs daily to export subscriber
data to our SFTP server
```
```
Fire a journey entry event for customer john@example.com
with the purchase data attached
```
## Metadata Skill
The Metadata skill deploys and retrieves Salesforce metadata. Use it for:
* **Apex** — Classes, triggers, test classes
* **Flows** — Screen flows, record-triggered flows, scheduled flows
* **LWC** — Lightning Web Components
* **Objects** — Custom objects and fields
* **Profiles & Permissions** — Profiles, permission sets, permission set groups
* **Other metadata** — Layouts, tabs, reports, dashboards, and 365+ types
**Example prompts:**
```
Deploy this Apex class to my sandbox
```
```
Retrieve all Flows from production so I can review them
```
## Combining Skills
Complex implementations often use multiple skills. The agent handles this automatically:
```
Create a unified customer profile in Data Cloud from Sales Cloud
and Marketing Cloud data, then create a segment for high-value
customers and activate it to an Agentforce service agent.
```
This request uses:
* **Data Cloud skill** — For connections, identity resolution, segments
* **Agentforce skill** — For the service agent configuration
The agent orchestrates across skills as needed.
## Next Steps
Learn how to guide the agent with natural language
See example prompts for building and deploying
# Using the Library
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
The **Library** is your workspace's shared home for [references](/concepts/references), [skills](/guides/using-skills), and collections. Open it from the sidebar.
## Find what you need
* **Mine** shows references you own or have starred.
* **Discover** shows everything shared with you — direct shares and anything published workspace-wide. It opens with a featured carousel and highlight rows.
Narrow the list with the **Types** filter (such as Recipes, Business Context, Brand Kits, and Documents) or the search box. The Library remembers the tab you were last on.
## Organize with collections
**Collections** group related references so you can attach them together.
* Open a collection from Discover to see everything inside it.
* Add a whole collection to an agent or workbench in one step, instead of attaching references individually.
## Skills
Skills show up in **Discover** as a **Skills** collection. Open it to browse the skills available to your workspace.
The Skills collection appears once your workspace has at least one skill. See [Creating Skills](/guides/creating-skills) to add your own.
## For admins
Workspace admins get a **Review** tab to approve or deny references that members request to publish, and can feature items and manage collections so the whole workspace sees the most useful context in Discover.
## Next Steps
Upload or import skills into the Library
Add and manage references
# Working with Workbenches
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
A [workbench](/docs/concepts/workbenches) groups multiple agents under one shared sandbox, with shared connections and references.
## Create one
From the dashboard drawer, click **Create workbench**, name it after the project, and attach the connections and references every agent in it should inherit.
## Add agents
The first agent shows as a tab in the strip up top. Click **+** to add more. Each new agent gets an auto-codename (Atlas, Echo, Nova…), its own chat history, and its own folder at `/workspace//`.
## When this pays off
Workbenches are most useful when you have **separate concerns running in parallel** that share the same connections. Some realistic shapes:
* **Domain split.** One agent owns sales pipeline analysis, another owns marketing performance. Both read from the same Salesforce + Marketing Cloud connections; neither has to context-switch.
* **Orchestrator + workers.** A scheduler-style agent triggers the others on a cron and rolls up their outputs into a weekly digest in `shared/`.
* **Long-lived ops space.** Same workbench, same references, fresh agents per question — keep months of history without one chat ballooning.
When a single agent's chat gets long, slow, or expensive, that's the signal to split it into a peer rather than keep going.
## Hand-offs
Agents in the same workbench message each other with the `peer.send_message` tool. Just say it in plain English:
> "When you're done, ask Echo to draft the follow-up email and post it to shared/."
Echo gets a structured peer message tagged with the sender's codename. Both sides of the conversation are visible — switch tabs to follow.
## Sharing files
```
/workspace/
├── / ← per-agent scratch
├── /
└── shared/ ← anything any agent should see
```
Agents already know this layout. "Save it to shared" or "look in shared for the brief" works. Workbench references live read-only at `/workspace/.refs//`.
## Editing connections and references mid-flight
Edit them on the workbench any time. Active agents pick up the change on their next turn — no restart.
## Next Steps
The concept reference
Single-agent basics
# Connecting
import { Callout } from 'fumadocs-ui/components/callout';
import { Tabs, Tab } from 'fumadocs-ui/components/tabs';
You need a MeshMesh workspace to connect — you sign in with the same account you use for the app, and your [role](/mcp/roles-and-access) decides which tools you get.
## Endpoint
```
POST https://gateway.meshmesh.io/mcp
```
The server speaks **Streamable HTTP** (MCP's HTTP transport). It is stateless — every request is a single `POST`. There is no `stdio` or SSE endpoint, and `GET`/`DELETE` return `405`.
## Authentication
The server uses **OAuth 2.0**. Your MCP client obtains a token through the standard authorization flow and sends it as a bearer token:
```
Authorization: Bearer
```
Clients discover the authorization server automatically — no manual setup beyond signing in:
* `GET https://gateway.meshmesh.io/.well-known/oauth-protected-resource`
* `GET https://gateway.meshmesh.io/.well-known/oauth-authorization-server`
A request with no token gets a `401` whose `WWW-Authenticate` header points at the discovery URL, which is the client's cue to start the OAuth flow. Your MeshMesh identity and workspace are taken from the token, so the tools you can see and call match your role (see [Roles & access](/mcp/roles-and-access)).
Most MCP clients handle OAuth for you: paste the URL, and the client opens a browser window to sign in. You do not create or paste an API key.
## Connecting from popular clients
Any MCP client that supports remote (HTTP) servers with OAuth works. On first use each client opens a browser to sign in; after that the model can call MeshMesh [tools](/mcp/tools). Pick your client below.
Add the server, then authenticate:
```bash
claude mcp add --transport http meshmesh https://gateway.meshmesh.io/mcp
```
In a Claude Code session, run `/mcp` and complete the browser sign-in. Verify with:
```bash
claude mcp list
```
Add `--scope user` (all your projects) or `--scope project` (shared via a checked-in `.mcp.json`) to change where the server is saved.
Add the server to `~/.codex/config.toml`:
```toml
[mcp_servers.meshmesh]
url = "https://gateway.meshmesh.io/mcp"
```
Then sign in:
```bash
codex mcp login meshmesh
```
Support for remote (HTTP) MCP servers with OAuth is recent in Codex — update to the latest CLI (`npm i -g @openai/codex@latest`) and check the Codex docs for the current `mcp` command if `codex mcp login` isn't recognized. Use `url` (not `command`/`args`, which are for local servers).
Add the server to `~/.cursor/mcp.json` (global) or `.cursor/mcp.json` (project):
```json
{
"mcpServers": {
"meshmesh": {
"url": "https://gateway.meshmesh.io/mcp"
}
}
}
```
Cursor detects that the server needs OAuth and prompts you to sign in — no token goes in the file.
Add the server to `.vscode/mcp.json`:
```json
{
"servers": {
"meshmesh": {
"type": "http",
"url": "https://gateway.meshmesh.io/mcp"
}
}
}
```
Or run **MCP: Add Server** from the Command Palette and paste the URL.
VS Code uses the key `servers` and requires `"type": "http"` — different from Cursor's `mcpServers`.
Connect through the UI:
1. **Settings → Connectors → Add custom connector**
2. Name it `meshmesh` and enter the URL `https://gateway.meshmesh.io/mcp`
3. Click **Add** and complete the browser sign-in.
MeshMesh tools then appear in the message composer's connectors menu.
Custom connectors require a Claude plan that supports them. If you don't see **Add custom connector**, your plan may not include it.
## Your first call
Once connected, confirm it works by asking your client to call `get_user_details` — it should return your email and workspace. Then kick off real work:
1. **`start_agent`** with a name and a prompt — this returns an `agentId`.
2. **`get_agent_status`** to watch progress and see when it's waiting on you.
3. **`get_plan`** then **`continue_agent`** to approve the plan or send a follow-up.
See the full [tool reference](/mcp/tools) for everything else.
## Troubleshooting
* **Sign-in loops or "needs authentication":** re-run the client's auth step (in Claude Code, `/mcp`). Tokens refresh automatically, but you can always clear and re-authenticate.
* **`405` errors:** the server only accepts `POST` at `/mcp` — make sure your client is configured for HTTP transport, not SSE.
* **A tool is missing:** it's likely admin-only and hidden for your role. See [Roles & access](/mcp/roles-and-access).
# MCP Server
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
MeshMesh exposes a **remote MCP server** so you can start and steer agents, manage references, and browse your Library from any [Model Context Protocol](https://modelcontextprotocol.io) client — Slack, Claude, or one you build yourself.
You connect your MCP client **to** MeshMesh; the client's model then calls MeshMesh tools on your behalf. Everything runs against your own workspace, scoped to your identity and role.
**Not the same as a Custom MCP.** Adding a *custom MCP* ([Adding Integrations](/guides/adding-integrations)) connects an outside MCP server **into** MeshMesh so your agents can use its tools. The MeshMesh MCP server is the opposite direction: it lets an **outside client drive MeshMesh**.
## What you can do
* **Run [agents](/concepts/agents)** — start one, approve its plan, send follow-ups, and switch between plan and build mode.
* **Track work** — check agent status, read plans, and open [artifacts](/concepts/artifacts).
* **Manage [references](/concepts/references) & connections** — create, update, and attach them to agents.
* **Browse the [Library](/concepts/library)** — search references, skills, and collections; admins can review and curate.
* **See what your team is doing** — a workspace-wide feed of recently completed agent runs (where enabled).
## Get started
Endpoint, authentication, and a copy-paste config
Every tool, what it does, and its parameters
Which tools your role can see and call
# Roles & Access
The MeshMesh MCP server enforces your workspace role on every request. Your role comes from your workspace membership, resolved from the identity in your OAuth token.
## Roles
Roles are ordered, least to most privileged:
**viewer → member → admin → owner**
## What each role sees
* **Any signed-in workspace user** gets the agent, artifact, reference, connection, workbench, and identity tools, plus Library **read** (`library_search`, `collection_list`).
* **Admins and owners** additionally see the Library **admin** tools — the review queue, publishing and curation, and collection management.
Admin-only tools are hidden entirely: if your role can't use a tool, it does not appear in your client's tool list and cannot be called. See the [Tool Reference](/mcp/tools) for which tools are admin-only.
# Tool Reference
import { Callout } from 'fumadocs-ui/components/callout';
The tools your client sees depend on your role. Tools marked **admin** are hidden unless you are a workspace admin or owner. `?` marks an optional parameter.
**Start here.** Most work is one loop: `start_agent` to kick off, `get_agent_status` to watch, `get_plan` / `continue_agent` to approve the plan and send follow-ups. The rest are for managing references, connections, and the Library.
## Agents & tasks
| Tool | What it does | Parameters |
| ------------------- | ------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------ |
| `start_agent` | Start a headless agent that builds in a Salesforce org; returns an `agentId` | `name`, `prompt`, `mode?` (plan / build, default build), `referenceIds?`, `connectionIds?` |
| `list_agents` | Compact index of your agents | `limit?` (1–100), `workbenchId?` |
| `continue_agent` | Send a follow-up message or approve a plan | `agentId`, `message`, `mode?` (plan / build) |
| `switch_agent_mode` | Flip between plan and build without a message | `agentId`, `mode` (plan / build) |
| `get_agent_status` | Current state, what it's waiting for, a summary, recent messages, and artifacts | `agentId` |
| `get_plan` | Show an agent's plan and its approval state | `agentId` |
| `edit_plan` | Revise the plan with an instruction; the agent re-plans | `agentId`, `instruction` |
Some agents can't be driven from MCP — for example ones started in an older version, or that use a desktop-only connection. `list_agents` flags these, and `continue_agent` explains why if you try.
## Artifacts
| Tool | What it does | Parameters |
| ---------------- | -------------------------------------------------------------------------------------------- | ------------ |
| `list_artifacts` | Compact index of an agent's artifacts | `agentId` |
| `get_artifact` | Render one artifact (Markdown, or an interactive widget for HTML if your client supports it) | `artifactId` |
## References
| Tool | What it does | Parameters |
| --------------------- | ----------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------- |
| `reference_get` | Get a reference by ID | `id` |
| `reference_list` | List your workspace references, or the references attached to one agent | `agentId?`, `category?`, `subType?`, `includePublic?` (default false), `limit?` (1–100, default 50), `offset?` |
| `reference_create` | Create a document reference from Markdown | `name`, `content`, `description?` |
| `reference_update` | Update a reference you own | `id`, `name?`, `description?`, `category?`, `subType?`, `metadata?` |
| `discover_references` | Keyword search across your own and shared-with-you references | `query`, `scope?` (own / shared / all, default all), `limit?` (1–100) |
Publishing a reference to your workspace isn't done over MCP — it goes through admin review (see the Library admin tools below). Ask an admin, or publish from the app.
## Attach & detach
| Tool | What it does | Parameters |
| ------------------- | ------------------------------------------------------------------ | ------------------------- |
| `attach_reference` | Attach a reference to an agent (takes effect live if it's running) | `agentId`, `referenceId` |
| `detach_reference` | Detach a reference from an agent | `agentId`, `referenceId` |
| `attach_connection` | Attach a connection, such as a Salesforce org | `agentId`, `connectionId` |
| `detach_connection` | Detach a connection from an agent | `agentId`, `connectionId` |
## Connections
| Tool | What it does | Parameters |
| ------------------ | ---------------------------------------------------------------------------- | -------------- |
| `list_connections` | List connections you can attach to an agent (server-stored connections only) | — |
| `get_connection` | One connection, plus the agents using it | `connectionId` |
## Workbenches
| Tool | What it does | Parameters |
| --------------------- | ----------------------------------------------------------------------------- | --------------------------------------------------------- |
| `list_workbenches` | Compact index of your workbenches | — |
| `get_workbench` | Full detail of one workbench | `workbenchId` |
| `create_workbench` | Create a workbench | `name`, `description?`, `connectionIds?`, `referenceIds?` |
| `get_recent_activity` | Overview of your workbenches, their agents, and any agents not in a workbench | `limit?` (1–200, default 25) |
## Identity
| Tool | What it does | Parameters |
| ------------------ | ------------------------------------------ | ---------- |
| `get_user_details` | Your identity — email, name, and workspace | — |
## Workspace activity
See what others in your workspace are doing — a team-wide feed of recently completed agent runs, not just your own.
| Tool | What it does | Parameters |
| ----------------------------- | ------------------------------------------------------- | --------------------------- |
| `get_workspace_activity_feed` | Recent completed agent runs across your whole workspace | `limit?` (1–100), `cursor?` |
Team activity isn't turned on for every workspace. If it's off for yours, this tool returns an error explaining that — reach out to your admin to enable it.
## Library — read
| Tool | What it does | Parameters |
| ----------------- | ----------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------- |
| `library_search` | Search the Library by scope (mine / starred / shared / discovery / featured / skills; admins also `review`) | `scope?` (default discovery), `search?`, `limit?` (1–100), `offset?` |
| `collection_list` | List Discovery collections, or the references in one collection | `search?`, `collectionId?`, `limit?` |
## Library — admin
These tools are only visible to workspace **admins** and **owners**.
| Tool | What it does | Parameters |
| ----------------------------- | ----------------------------------------------------- | ---------------------------------------------------------- |
| `review_queue_list` | List references pending publish review | `limit?`, `offset?` |
| `reference_review` | Approve or deny a publish request | `requestId`, `decision` (approve / deny), `note?` |
| `reference_set_featured` | Set or clear a reference's featured rank in Discovery | `referenceId`, `featuredRank` (0–100000, or null to clear) |
| `reference_publish_admin` | Publish a reference directly, skipping review | `referenceId` |
| `reference_update_admin` | Edit any reference's title, description, or tags | `referenceId`, `name?`, `description?`, `tags?` |
| `collection_create` | Create a Discovery collection | `name`, `description?` |
| `collection_add_references` | Add references to a collection | `collectionId`, `referenceIds` (1–100) |
| `collection_remove_reference` | Remove a reference from a collection | `collectionId`, `referenceId` |
# Commerce Cloud
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
import { Tabs, Tab } from 'fumadocs-ui/components/tabs';
import { Accordion, Accordions } from 'fumadocs-ui/components/accordion';
# Commerce Cloud
Commerce Cloud is Salesforce's e-commerce platform for B2C and B2B digital experiences. Build mobile-first storefronts with AI-powered personalization and seamless Salesforce integration.
B[Product Catalog]
B --> C[AI Recommendations]
C --> D[Add to Cart]
D --> E[Checkout]
E --> F{Payment Success?}
F -->|Yes| G[Order Fulfillment]
F -->|No| H[Retry Payment]
H --> E
G --> I[Customer Success]
`}
/>
## What You Can Do
Mobile-first storefronts with personalized product recommendations
Account-based pricing, bulk ordering, and approval workflows
Real-time inventory, fulfillment tracking, and returns processing
Multiple gateways, subscriptions, and international currencies
## Product Catalog Management
Manage products, categories, pricing, and inventory.
Ask MeshMesh: "Show me all products with low inventory that are on sale"
**Common Use Cases:**
* Product information management
* Category organization and navigation
* Variant management (size, color, style)
* Pricing and promotions
**Example Request:**
"Update inventory for SKU12345 to 50 units and apply a 20% discount"
## Shopping Experience
Create personalized shopping journeys with AI-powered recommendations.
Intelligent search, faceted filtering, and category browsing. Einstein recommendations suggest products based on browsing and purchase history.
Dynamic content based on customer behavior, location, and preferences. Personalized pricing for B2B customers and loyalty tiers.
One-page checkout optimization. Multiple payment methods, saved addresses, and express checkout for returning customers.
## API Operations
MeshMesh handles API complexity automatically. Describe what you need in plain English.
```http
GET /s/{site_id}/dw/shop/v21_3/products
?q={search_query}
&category_id={category}
&limit=50
```
Retrieve products, search, and filter by category or attributes.
```json
POST /baskets/{basket_id}/items
{
"product_id": "prod_123",
"quantity": 2
}
```
Add items to cart, update quantities, apply coupons, and calculate totals.
```http
POST /orders
```
Create orders, process payments, and track order status. Retrieve order history and details.
## Key Objects
**Product** stores item information and attributes.
**Fields:**
* id, name, description
* price, currency
* categories, variations
* images, inventory
**Example:** Manage product catalog with pricing and stock levels.
**Order** tracks customer purchases.
**Fields:**
* order\_no, status, total
* customer\_info, billing, shipping
* product\_items, payments
* shipments, tracking
**Example:** Process and fulfill customer orders.
**Customer** maintains shopper profiles.
**Fields:**
* customer\_no, email
* addresses, payment\_methods
* order\_history, preferences
* loyalty\_points
**Example:** Personalize experiences based on customer data.
**Inventory** manages stock across locations.
**Fields:**
* product\_id, location\_id
* quantity\_on\_hand
* allocated, available
* reorder\_threshold
**Example:** Track stock levels and fulfill from optimal locations.
## Use Cases
Unified inventory, buy online pick up in-store, and consistent pricing
Progressive web apps, mobile checkout, and location-based services
Multi-language, multi-currency, and region-specific tax calculations
Account-based catalogs, contract pricing, and approval workflows
## Authentication
Commerce Cloud uses OAuth 2.0 for API authentication.
Use for server-to-server integrations and backend operations.
```http
POST /oauth2/token
Content-Type: application/x-www-form-urlencoded
grant_type=client_credentials&
client_id=YOUR_CLIENT_ID&
client_secret=YOUR_CLIENT_SECRET
```
Use for customer-facing operations and personalized experiences.
Implement guest checkout or registered customer authentication based on your business needs.
## Tips and Tricks
**Product Management:**
* Use product sets for bundling
* Optimize images for mobile
* Enable product recommendations
* Set up inventory alerts
**Performance Optimization:**
* Cache product catalog data
* Use CDN for static assets
* Optimize search indexing
* Monitor page load times
**Conversion Optimization:**
* Simplify checkout process
* Enable guest checkout
* Show trust badges
* Offer multiple payment options
## Next Steps
Set up your Commerce Cloud connection
Learn more about managing products
Explore e-commerce scenarios
# Content Management
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
import { Tabs, Tab } from 'fumadocs-ui/components/tabs';
import { Steps } from 'fumadocs-ui/components/steps';
# Content Management
Salesforce Content Management provides centralized storage and distribution for digital assets across all Salesforce products. Manage images, videos, documents, and marketing content with enterprise-grade security.
B[Content Management]
B --> C[Marketing Cloud]
B --> D[Experience Cloud]
B --> E[Commerce Cloud]
B --> F[Service Cloud]
B --> G[Sales Cloud]
C --> H[Customer Experiences]
D --> H
E --> H
F --> H
G --> H
`}
/>
## What You Can Do
Store and organize images, videos, documents, and files
Create reusable content blocks and email templates
Track asset changes and maintain version history
Share content seamlessly across Salesforce products
## Digital Asset Management
Centralize all your marketing and sales assets in one secure location.
Ask MeshMesh: "Upload these product images and organize them by category"
**Supported File Types:**
* Images: JPEG, PNG, GIF, WebP, SVG
* Videos: MP4, MOV, AVI, WebM
* Documents: PDF, DOC, PPT, XLS
* Code: HTML, CSS, JavaScript
**Example Request:**
"Find all unused marketing assets from last quarter and archive them"
## Content Builder
Create and manage marketing content with drag-and-drop tools.
Build responsive email templates with reusable components. Save brand-compliant layouts for consistent messaging.
Create modular content pieces used across campaigns. Update once, change everywhere automatically.
Design conversion-optimized landing pages. Integrate with forms and tracking for campaign measurement.
## Organize Content
### Create Folder Structure
Organize assets by campaign, product line, or content type for easy discovery.
### Add Metadata Tags
Tag assets with keywords, categories, and properties for powerful search capabilities.
### Set Permissions
Control who can view, edit, or delete content with role-based access controls.
### Enable Version Control
Track changes, revert to previous versions, and maintain asset history automatically.
## API Operations
MeshMesh handles API complexity. Just describe what you need in natural language.
```json
POST /asset/v1/content/assets
Content-Type: multipart/form-data
{
"name": "product-hero-image",
"assetType": {
"name": "image",
"id": 28
},
"file": [binary data]
}
```
```http
GET /asset/v1/content/assets
?$filter=name%20like%20'product%'
&$orderby=modifiedDate%20desc
```
Search by name, tags, type, or date. Filter and sort results.
```json
POST /asset/v1/content/assets
{
"name": "Welcome Email Header",
"assetType": {"name": "htmlblock"},
"content": "
Welcome!
"
}
```
## Cross-Cloud Integration
Content Management integrates seamlessly with all Salesforce products.
Email templates, images, and journey content blocks
Community assets, knowledge articles, and portal branding
Product images, category banners, and promotional content
Knowledge base articles, support docs, and training materials
Presentation templates, product collateral, and case studies
Dashboard images, report templates, and visualization assets
## Content Workflows
Automate content approval and publishing with built-in workflows.
Create content in draft mode. Collaborate with team members and make revisions before submission.
Submit for review when ready. Reviewers receive notifications and can add comments.
Approvers accept or reject with feedback. Multi-level approval chains supported for compliance.
Approved content publishes automatically. Schedule publication dates and times in advance.
## Security and Governance
Encryption at rest and in transit. Role-based access controls and complete audit trails.
**Access Controls:**
* Role-based permissions
* Content approval workflows
* Brand compliance enforcement
* Audit trails for all changes
**Data Protection:**
* Encryption at rest and in transit
* Geographic data residency
* Automated PII detection
* Content lifecycle management
## Tips and Tricks
**Asset Organization:**
* Use consistent naming conventions
* Tag assets with searchable metadata
* Create logical folder hierarchies
* Archive outdated content regularly
**Performance Optimization:**
* Optimize image sizes for web
* Use CDN for global delivery
* Enable caching where appropriate
* Monitor storage usage
**Content Reuse:**
* Build modular content blocks
* Create template libraries
* Use dynamic content tokens
* Share assets across teams
## Next Steps
Set up your Content Management connection
Learn more about organizing content
Explore content management scenarios
# Data Cloud
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
import { Tabs, Tab } from 'fumadocs-ui/components/tabs';
import { Steps } from 'fumadocs-ui/components/steps';
# Data Cloud
Data Cloud is Salesforce's Customer Data Platform (CDP). Unify customer data from multiple sources, create 360-degree profiles, and activate data across all your systems in real-time.
B[Data Ingestion]
B --> C[Identity Resolution]
C --> D[Unified Customer Profile]
D --> E[Segmentation Engine]
E --> F[Marketing Cloud]
E --> G[Sales Cloud]
E --> H[Service Cloud]
E --> I[Commerce Cloud]
`}
/>
## What You Can Do
Merge data from multiple sources into single customer profiles
Create real-time audiences based on any customer attribute or behavior
Send segments to Marketing, Sales, Service, and Commerce clouds
Use AI to predict churn, lifetime value, and next best actions
## Unified Customer Profiles
Create a single source of truth for each customer by merging data from all touchpoints.
Ask MeshMesh: "Show me high-value customers who purchased in the last 30 days"
**Identity Resolution:**
* Match customers across email, phone, and IDs
* Configure matching rules and thresholds
* Resolve duplicates automatically
* Maintain data lineage
**Example Request:**
"Create a unified profile merging web, mobile, and in-store purchase data"
## Real-Time Segmentation
Build dynamic audience segments that update in real-time as customer behavior changes.
Segment based on actions: purchases, page views, email clicks, support cases. Updates instantly as customers interact.
Traditional segmentation: age, location, income, industry. Combine with behavioral data for precision targeting.
AI-powered segments: likely to churn, high lifetime value, next best product. Einstein predictions built-in.
## Data Ingestion
### Connect Data Sources
Link Salesforce clouds, external databases, data warehouses, and third-party apps.
### Map Data Fields
Define how source fields map to Data Cloud's standard data model objects.
### Configure Sync
Set up real-time streaming or scheduled batch imports based on data volume and freshness needs.
### Validate Data
Monitor data quality, completeness, and ingestion errors through dashboards.
## Query with SQL
Query unified customer profiles using SQL syntax.
MeshMesh converts natural language to SQL automatically. No need to write queries manually.
```sql
SELECT
UnifiedIndividual.Id,
UnifiedIndividual.FirstName__c,
UnifiedIndividual.Email__c,
UnifiedIndividual.LifetimeValue__c
FROM UnifiedIndividual__dlm
WHERE UnifiedIndividual.LifetimeValue__c > 1000
AND UnifiedIndividual.LastPurchaseDate__c >= LAST_N_DAYS:30
```
## Data Activation
Send segments to Salesforce clouds and external systems for immediate use.
Personalized email campaigns and journey orchestration
Intelligent lead scoring and account insights
Context-aware customer support with complete history
Dynamic pricing and product recommendations
## Key Objects
**Unified Individual** combines all customer data into one profile.
**Standard Fields:**
* FirstName\_\_c, LastName\_\_c
* Email\_\_c, Phone\_\_c
* CreatedDate\_\_c
**Calculated Fields:**
* LifetimeValue\_\_c (total customer value)
* LastPurchaseDate\_\_c (most recent transaction)
* PreferredChannel\_\_c (primary interaction channel)
**Data Model Objects (DMO)** standardize business data structures.
**Individual:** Person demographics and contact info
**Engagement:** Interactions, campaigns, touchpoints
**Purchase:** Transactions, orders, products
**Custom:** Industry-specific data models
**Segment** defines audience criteria and membership.
**Properties:**
* Name, description, criteria
* Member count, refresh schedule
* Activation targets
* Creation and update timestamps
## API Operations
```json
POST /connect/insights/
{
"records": [{
"timestamp": "2024-03-15T10:30:00Z",
"eventType": "Purchase",
"customerId": "cust_12345",
"attributes": {
"productId": "prod_abc",
"amount": 99.99
}
}]
}
```
Use SQL to query unified profiles, calculated insights, and segment membership through the Query API.
Create, update, and activate segments programmatically. Monitor segment size and refresh status.
## Privacy and Compliance
Data Cloud includes tools for GDPR, CCPA, and other privacy regulations.
**Privacy Features:**
* Right to be forgotten (data deletion)
* Consent management and tracking
* Data minimization with auto-delete
* Field-level data classification
* Complete audit trails
**Example Request:**
"Delete all data for [customer@example.com](mailto:customer@example.com) and log the deletion"
## Tips and Tricks
**Data Quality:**
* Set up data validation rules
* Monitor ingestion error rates
* Configure matching thresholds carefully
* Review unified profiles regularly
**Segmentation Best Practices:**
* Start with simple segments
* Test before activating to production
* Monitor segment size changes
* Use calculated insights for efficiency
**Performance Optimization:**
* Use batch ingestion for historical data
* Stream real-time events for immediacy
* Index frequently queried fields
* Archive old engagement data
## Next Steps
Set up your Data Cloud connection
Learn more about querying unified profiles
Explore customer data platform scenarios
# Einstein Analytics
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
import { Tabs, Tab } from 'fumadocs-ui/components/tabs';
import { Accordion, Accordions } from 'fumadocs-ui/components/accordion';
# Einstein Analytics
Einstein Analytics (now part of Tableau CRM) is Salesforce's AI-powered business intelligence platform. Create interactive dashboards, predictive analytics, and machine learning models to make data-driven decisions.
B[Data Integration]
C[External Data] --> B
B --> D[Dataset Creation]
D --> E[Analytics Engine]
E --> F[Interactive Dashboards]
E --> G[Einstein AI Insights]
G --> H[Predictive Models]
F --> I[Business Decisions]
H --> I
`}
/>
## What You Can Do
Build visual analytics with charts, graphs, and real-time data
Get automated insights, predictions, and anomaly detection
Connect Salesforce data and external sources seamlessly
Forecast outcomes and identify trends with machine learning
## Data Visualization
Create interactive dashboards that tell compelling data stories.
Ask MeshMesh: "Show me sales performance by region for Q4 in a dashboard"
**Visualization Types:**
* Charts: Bar, line, pie, scatter, funnel
* Tables: Sortable, filterable data grids
* Maps: Geographic data visualization
* Custom: Build unique visualizations
**Example Request:**
"Create a dashboard showing revenue trends, pipeline analysis, and win rates"
## Einstein AI Insights
Get automated insights and predictions powered by machine learning.
Einstein Discovery analyzes your data and surfaces key insights automatically. Explains patterns, correlations, and drivers behind outcomes.
Build predictive models without coding. Forecast churn, lifetime value, deal closure, and more with AI explanations.
Automatic anomaly detection alerts you to unusual patterns. Catch issues early and investigate root causes.
## SAQL Query Language
Query datasets using SAQL (Salesforce Analytics Query Language).
MeshMesh converts natural language to SAQL automatically. No need to memorize syntax.
```sql
q = load "opportunity_dataset";
q = filter q by 'StageName' == "Closed Won";
q = group q by 'Owner.Name';
q = foreach q generate
'Owner.Name' as 'Owner',
sum('Amount') as 'Total_Revenue';
q = order q by 'Total_Revenue' desc;
q = limit q 10;
```
Top 10 sales reps by revenue from won opportunities.
```sql
q = load "opportunity_dataset";
q = foreach q generate
'Name',
'Amount',
'CloseDate',
days_between('CreatedDate', 'CloseDate') as 'SalesCycle';
q = filter q by 'SalesCycle' < 90;
q = order q by 'Amount' desc;
```
Opportunities closed within 90 days, sorted by amount.
Common aggregate functions in SAQL:
* `sum(field)` - Total values
* `avg(field)` - Average calculation
* `count()` - Count records
* `min(field)` / `max(field)` - Minimum/Maximum
* `unique(field)` - Distinct values
## API Operations
```http
GET /services/data/v60.0/wave/datasets
?q={search_term}
&sort=CreatedDate&order=desc
```
List, create, update, and delete datasets. Upload CSV data or connect to external sources.
```http
GET /wave/dashboards/{dashboard_id}
?embed=true&format=json
```
Access dashboard definitions, widgets, and configurations programmatically.
```http
POST /wave/query
{
"query": "q = load \"dataset\"; q = limit q 100;"
}
```
Execute SAQL queries and retrieve results via API.
## Use Cases
Revenue forecasting, pipeline analysis, and rep performance tracking
Multi-touch attribution, campaign ROI, and lead source analysis
Lifetime value analysis, churn prediction, and behavioral segmentation
Service case analysis, team productivity, and process optimization
## Key Objects
**Dataset** stores analytics data collections.
**Fields:**
* id, name, label
* rowCount (number of records)
* fields (column definitions)
* datasetType (source)
**Example:** Import opportunity data for sales analysis.
**Dashboard** defines interactive visualizations.
**Fields:**
* id, name, description
* folderId (parent folder)
* state (configuration JSON)
* layout (grid positioning)
**Example:** Create executive sales dashboard.
**Lens** represents saved explorations and charts.
**Fields:**
* id, name, type
* datasetId (source dataset)
* visualizationType
* query (SAQL)
**Example:** Save frequently used sales report view.
## Einstein Discovery
Build predictive models with automated machine learning.
Einstein Discovery builds ML models from your data. Just define your goal and let AI do the work.
**What You Can Predict:**
* Customer churn probability
* Deal win likelihood
* Lifetime customer value
* Product recommendations
* Next best action
**How It Works:**
1. Select dataset and outcome variable
2. Einstein analyzes patterns automatically
3. Review AI-generated insights and model
4. Deploy predictions to Salesforce
5. Monitor and improve over time
## Tips and Tricks
**Dashboard Design:**
* Start with key metrics at the top
* Use consistent color schemes
* Limit dashboards to 5-7 widgets
* Add filters for interactivity
**Query Performance:**
* Filter data early in SAQL queries
* Use appropriate aggregations
* Limit result sets when possible
* Index frequently queried fields
**Data Quality:**
* Clean source data before import
* Validate data types and formats
* Set up automated data refresh
* Monitor dataset health
## Next Steps
Set up your Einstein Analytics connection
Learn more about querying datasets
Explore analytics and BI scenarios
# Experience Cloud
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
import { Tabs, Tab } from 'fumadocs-ui/components/tabs';
import { Steps } from 'fumadocs-ui/components/steps';
# Experience Cloud
Experience Cloud lets you build branded digital experiences for customers, partners, and employees. Create communities, portals, and websites with seamless Salesforce integration.
>Community: Visit self-service portal
Community->>Knowledge: Search for solution
alt Solution Found
Knowledge-->>Customer: Display article
Customer->>Customer: Resolve issue
else Needs Help
Customer->>Support: Create case
Support->>Support: Agent reviews
Support-->>Customer: Provide solution
Customer->>Community: Rate article
end
`}
/>
## What You Can Do
Self-service portals with knowledge base and peer support
Collaboration spaces for partners with deal registration
Internal communication and knowledge sharing
Branded websites with Salesforce data integration
## Community Management
Build engaging communities for different audiences.
Ask MeshMesh: "Show me the most active community members this month"
**Key Features:**
* Branded community sites
* User-generated content
* Gamification (badges, points)
* Moderation workflows
**Example Request:**
"Create a new customer community with knowledge base and case creation"
## Community Types
Self-service support with knowledge articles, case tracking, and peer-to-peer forums. Reduce support costs with deflection.
Dedicated partner collaboration with deal registration, training materials, and resource libraries. Enable partner success.
Internal communication hub with company news, knowledge sharing, and employee recognition programs.
## Knowledge Base Integration
Integrate Salesforce knowledge articles for self-service support.
### Create Knowledge Articles
Write help documentation, FAQs, and troubleshooting guides in Salesforce.
### Configure Search
Enable powerful search across articles. AI recommends relevant content automatically.
### Add to Community
Display articles in your community with categories and featured content.
### Track Effectiveness
Monitor which articles resolve cases. Update based on customer feedback.
## API Operations
MeshMesh handles API complexity. Describe community operations in natural language.
```http
GET /connect/communities/{community_id}/members
?pageParam={page_token}&pageSize=25
```
Manage community members, roles, and permissions.
```json
POST /connect/communities/{community_id}/chatter/feed-elements
{
"body": {
"messageSegments": [{
"type": "Text",
"text": "Hello Community!"
}]
}
}
```
Create posts, discussions, and announcements.
Track community engagement, popular content, and member activity. Export metrics for reporting.
## Experience Builder
Design and customize your community with drag-and-drop tools.
Drag-and-drop components to create custom pages
Add Lightning Web Components for advanced features
Automatic mobile optimization for all devices
Meta tags, sitemaps, and search engine optimization
## Gamification
Engage members with points, badges, and leaderboards.
Award reputation points for helpful answers, posts, and participation. Members earn status as they contribute.
Create custom badges for achievements. Recognize experts, helpful members, and milestones.
Display top contributors and active members. Foster friendly competition and engagement.
## Key Objects
**Network** represents the community configuration.
**Fields:**
* Id, Name, UrlPathPrefix
* Status (Active/Inactive)
* Template (site template type)
**Example:** Main community or site settings.
**NetworkMember** represents user profiles.
**Fields:**
* NetworkId, MemberId
* ReputationPoints
* DefaultGroupNotificationFrequency
* Profile, Role
**Example:** Community user accounts and permissions.
**FeedElement** represents posts and content.
**Fields:**
* Type (TextPost, ContentPost, LinkPost)
* Body, Title
* CreatedDate, Visibility
* Likes, Comments
**Example:** Community posts and discussions.
## Use Cases
Self-service knowledge base reduces case volume by 30-50%
Training, resources, and deal registration in one portal
User-generated reviews, feature requests, and beta testing
Internal communication, recognition, and knowledge sharing
## Authentication and Security
Control community access with flexible authentication options.
Use role-based permissions, data security rules, and guest user profiles appropriately.
**Authentication Methods:**
* Salesforce login
* Social login (Google, Facebook)
* Single sign-on (SSO)
* Self-registration
**Security Controls:**
* Profile-based permissions
* Sharing rules for data access
* Content moderation
* Guest user restrictions
## Tips and Tricks
**Community Engagement:**
* Seed initial content before launch
* Feature top contributors
* Respond to questions quickly
* Run contests and challenges
**Performance Optimization:**
* Enable caching for static content
* Optimize images and media
* Use Lightning components efficiently
* Monitor page load times
**Content Strategy:**
* Organize knowledge by topic
* Create featured articles
* Update content regularly
* Encourage user contributions
## Next Steps
Set up your Experience Cloud connection
Learn more about building communities
Explore community and portal scenarios
# Marketing Cloud Engagement
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
import { Tabs, Tab } from 'fumadocs-ui/components/tabs';
import { Steps } from 'fumadocs-ui/components/steps';
# Marketing Cloud Engagement
Marketing Cloud Engagement is Salesforce's comprehensive email marketing and customer journey platform. Create personalized, data-driven campaigns across email, mobile, social, web, and advertising channels.
B[Journey Builder]
B --> C[Email Studio]
B --> D[Mobile Studio]
B --> E[Social Studio]
C --> F[Personalized Customer Experience]
D --> F
E --> F
`}
/>
## What You Can Do
Build responsive emails with drag-and-drop tools, A/B testing, and send optimization
Orchestrate multi-channel campaigns with conditional logic and behavior-based automation
Send SMS/MMS messages and push notifications for mobile app engagement
Store and manage customer data with custom tables and SQL queries
## Email Campaigns
Use Email Studio to create and send targeted email campaigns.
Skip the API complexity. Just describe what you want: "Send a welcome email to all new subscribers from last week"
**Common Use Cases:**
* Welcome series for new subscribers
* Abandoned cart recovery
* Re-engagement campaigns
* Birthday and anniversary emails
**Example Request:**
"Create an email template for our summer sale and send it to VIP customers"
## Customer Journeys
Journey Builder orchestrates multi-channel campaigns that respond to customer behavior in real-time.
Trigger emails based on customer actions, time delays, or data changes. Perfect for onboarding sequences and nurture campaigns.
Combine SMS, push notifications, and in-app messages. Ideal for time-sensitive promotions and transactional updates.
Coordinate across email, mobile, social, and advertising. Deliver consistent messaging across all touchpoints.
## Data Management
Data Extensions store your customer information and campaign data.
**Key Capabilities:**
* Custom data tables with flexible schemas
* SQL query definitions for segmentation
* Import definitions for batch data loads
* API access for real-time updates
**Example Request:**
"Show me all customers who purchased in the last 30 days from my CustomerData extension"
## AMPScript Automation
AMPScript adds dynamic personalization to your emails.
MeshMesh generates AMPScript automatically. Describe your personalization needs in plain English.
```javascript
%%=Concat('Hello ', ProperCase(FirstName), '!')=%%
```
Greet customers by name with proper capitalization.
```javascript
%%=Lookup('CustomerData', 'Points', 'Email', emailaddr)=%%
```
Pull customer loyalty points from a data extension.
```javascript
%%[
IF @tierLevel == "VIP" THEN
SET @discount = "25%"
ELSE
SET @discount = "10%"
ENDIF
]%%
```
Show different discounts based on customer tier.
## Authentication
### Register Your Application
Create an app in Marketing Cloud with API access permissions.
### Request Access Token
Use OAuth 2.0 client credentials flow to get your token.
```http
POST /v2/token
Content-Type: application/json
{
"grant_type": "client_credentials",
"client_id": "YOUR_CLIENT_ID",
"client_secret": "YOUR_CLIENT_SECRET"
}
```
### Use Token in API Calls
Include the token in your request headers.
## Common API Operations
Create, update, and query custom data tables
Trigger transactional or promotional email sends
Access open, click, and bounce data
Update contact preferences and list memberships
## Security
* Rotate API tokens regularly
* Use OAuth 2.0 for all integrations
* Enable IP allowlisting when possible
* Monitor rate limits: 2,500 requests/hour
**Data Protection:**
* Field-level encryption for sensitive data
* GDPR and CCPA compliance features
* Automated data retention policies
* Role-based access controls
## Tips and Tricks
**Optimize Email Performance:**
* Use A/B testing for subject lines and content
* Segment audiences for better targeting
* Monitor engagement metrics and adjust
* Keep email lists clean and up-to-date
**AMPScript Best Practices:**
* Minimize data extension lookups
* Cache lookup results in variables
* Validate data before using it
* Comment complex logic for maintenance
**Journey Builder Tips:**
* Start simple and add complexity gradually
* Test journeys with small audiences first
* Use wait activities strategically
* Monitor journey analytics regularly
## Next Steps
Set up your Marketing Cloud Engagement connection
Explore common marketing automation scenarios
Learn about secure integration practices
# Marketing Cloud Next
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
import { Tabs, Tab } from 'fumadocs-ui/components/tabs';
import { Accordion, Accordions } from 'fumadocs-ui/components/accordion';
# Marketing Cloud Next
Marketing Cloud Next represents the future of AI-powered marketing automation. Built on autonomous agents and intelligent workflows, it delivers personalized experiences at scale through self-optimizing campaigns.
B[Marketing Agent]
B --> C{AI Analysis}
C --> D[Audience Discovery]
C --> E[Content Generation]
C --> F[Channel Selection]
D --> G[Campaign Launch]
E --> G
F --> G
G --> H[Performance Monitoring]
H --> I{Meets Goals?}
I -->|No| J[Auto-Optimize]
J --> G
I -->|Yes| K[Success]
`}
/>
## What You Can Do
Autonomous agents that create, optimize, and manage campaigns
Predictive analytics, content generation, and send-time optimization
Cross-channel coordination with real-time personalization
Self-improving campaigns that learn from performance
## Autonomous Marketing Agents
Deploy AI agents that handle complex marketing tasks with minimal oversight.
Ask MeshMesh: "Create an agent to optimize my Black Friday campaign with a $50k budget"
**Agent Capabilities:**
* Goal setting and KPI definition
* Audience discovery and segmentation
* Creative generation and testing
* Channel selection and timing
* Performance monitoring and optimization
**Example Request:**
"Deploy a campaign agent to increase Q4 conversion rate by 15% within budget"
## Agent Types
**Autonomous campaign creation and management**
Sets goals, discovers audiences, generates content, selects channels, and optimizes performance automatically. Minimal human intervention required.
**Individual customer experience optimization**
Analyzes behavior, matches content, optimizes timing, learns channel preferences, and manages lifecycle progression for each customer.
**Campaign performance improvement**
Monitors metrics, tests variations, allocates budget, adjusts targeting, and scales winning strategies in real-time.
## Einstein AI Integration
Built-in AI powers intelligent decisions across all marketing operations.
AI identifies high-value segments, predicts customer behavior, and recommends optimal audiences for campaigns automatically.
Generate email subject lines, body copy, social media posts, and landing pages. Einstein creates variations and tests automatically.
Predict the best time to reach each individual customer. Einstein learns from engagement patterns to maximize open and click rates.
Agents run continuous experiments on content, timing, and audiences. Winners scale automatically while losers are removed.
## API Operations
MeshMesh converts natural language to agent configurations and API calls automatically.
```json
POST /v1/agents
{
"name": "Q4_Campaign_Agent",
"type": "campaign_optimizer",
"objectives": {
"primary": "conversion_rate",
"target": 0.15
},
"constraints": {
"budget": 50000,
"duration": "90d"
}
}
```
```http
GET /v1/agents/{agent_id}/metrics
?timeframe=last_30_days
&granularity=daily
```
Track agent decisions, performance trends, and optimization actions taken.
```http
GET /einstein/v1/predictions
?model=customer_lifetime_value
&audience_id=segment_12345
&prediction_window=90d
```
Access Einstein AI predictions and recommendations.
## Key Objects
**Agent** represents autonomous marketing automation.
**Properties:**
* id, name, type, status
* objectives (goals and KPIs)
* constraints (budget, time, compliance)
* learning\_rate (optimization speed)
**Example:** Campaign optimizer agent for holiday promotions.
**Prediction** stores AI-generated insights.
**Properties:**
* prediction\_id, model\_type
* confidence\_score
* prediction\_value
* feature\_importance
**Example:** Customer churn probability prediction.
**Journey** orchestrates multi-channel experiences.
**Properties:**
* journey\_id, agent\_id
* dynamic\_paths (AI-controlled)
* optimization\_metrics
* customer\_count
**Example:** AI-managed customer onboarding journey.
## Use Cases
Self-optimizing email campaigns with dynamic subject lines and budget reallocation
AI-predicted next best actions with dynamic path adjustments
AI-generated emails, social posts, and landing pages at scale
Individual send-time optimization and contextual messaging
## Agent Configuration
Begin with single-objective agents. Add complexity as you learn what works for your business.
**Define Objectives:**
* Primary goal (conversion, engagement, revenue)
* Target metric value
* Success criteria
**Set Constraints:**
* Budget limits and allocation rules
* Time duration and deadlines
* Compliance requirements
* Brand guidelines
**Configure Learning:**
* Optimization aggressiveness
* Intervention thresholds
* Human override triggers
* Testing parameters
## Migration from Marketing Cloud Engagement
Marketing Cloud Next builds on Engagement foundations while adding autonomous capabilities.
* Journey Builder → Autonomous Journey Agents
* Email Studio → AI Content Generation
* Automation Studio → Intelligent Orchestration
* Analytics → Predictive Insights
Existing data extensions, AMPScript, and API endpoints remain functional. Progressive feature adoption supported.
Start by adding AI features to existing campaigns. Gradually transition to agent-based automation at your own pace.
## Tips and Tricks
**Agent Management:**
* Start with limited budgets while learning
* Monitor agent decisions closely initially
* Set conservative intervention thresholds
* Review performance weekly
**Optimization Strategy:**
* Define clear, measurable objectives
* Allow sufficient data for learning
* Don't over-constrain agents initially
* Balance automation with oversight
**AI Best Practices:**
* Provide quality training data
* Validate predictions before full deployment
* Monitor for bias and drift
* Update models regularly
## Next Steps
Set up your Marketing Cloud Next connection
Learn how to configure marketing agents
Explore autonomous marketing scenarios
# Sales Cloud
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
import { Tabs, Tab } from 'fumadocs-ui/components/tabs';
import { Accordion, Accordions } from 'fumadocs-ui/components/accordion';
# Sales Cloud
Sales Cloud is Salesforce's CRM platform for managing leads, opportunities, accounts, and customer relationships. Automate your sales process from lead generation to deal closure.
B{Qualify?}
B -->|Yes| C[Convert to Opportunity]
B -->|No| D[Nurture with Automation]
C --> E[Manage Pipeline]
E --> F{Close Deal?}
F -->|Yes| G[Customer Success]
F -->|No| H[Update & Follow Up]
H --> E
`}
/>
## What You Can Do
Capture, score, route, and convert leads automatically
Manage your sales pipeline with visual forecasting
Maintain complete customer relationship histories
Automate tasks, emails, and workflows
## Lead Management
Capture and qualify leads efficiently with automated routing and scoring.
Ask MeshMesh: "Show me all hot leads from this month that haven't been contacted yet"
**Common Use Cases:**
* Web-to-lead form creation
* Behavioral and demographic scoring
* Territory-based routing
* Automated follow-up sequences
**Example Request:**
"Create a new lead for John Smith at Acme Corp with email [john@acme.com](mailto:john@acme.com)"
## Opportunity Pipeline
Track deals through your sales process with stage-by-stage management.
Visualize all opportunities by stage, owner, and close date. Quickly identify bottlenecks and at-risk deals.
Predict revenue with rollup hierarchies and probability-weighted forecasting across your team.
Track competitors, deal outcomes, and lessons learned to improve your sales process.
## SOQL Queries
Query Salesforce data using SOQL (Salesforce Object Query Language).
MeshMesh converts natural language to SOQL automatically. No need to write queries manually.
```sql
SELECT Id, Name, Email, Company, Status
FROM Lead
WHERE CreatedDate = THIS_MONTH
AND Status IN ('Open - Not Contacted', 'Working - Contacted')
ORDER BY CreatedDate DESC
LIMIT 100
```
```sql
SELECT StageName, AVG(Amount), COUNT(Id)
FROM Opportunity
WHERE CloseDate >= THIS_QUARTER
GROUP BY StageName
ORDER BY AVG(Amount) DESC
```
```sql
SELECT Id, Name, Email,
(SELECT Id, Subject FROM Tasks),
(SELECT Id, Subject FROM Events)
FROM Lead
WHERE CreatedDate = THIS_MONTH
```
## Key Objects
**Lead** represents potential customers before qualification.
**Standard Fields:**
* FirstName, LastName, Email, Phone
* Company, Status, Rating
* LeadSource, Industry
**Example:** Track incoming prospects and their qualification status.
**Opportunity** represents potential sales deals.
**Standard Fields:**
* Name, AccountId, Amount
* StageName, CloseDate, Probability
* Type, LeadSource
**Related Objects:** OpportunityLineItem, OpportunityContactRole, OpportunityHistory
**Account** represents companies or organizations.
**Standard Fields:**
* Name, Type, Industry
* Phone, Website, BillingAddress
* ParentId (for hierarchies)
**Example:** Maintain complete customer and prospect company records.
**Contact** represents individuals at accounts.
**Standard Fields:**
* FirstName, LastName, Email
* AccountId, Title, Phone
* ReportsToId
**Example:** Track decision-makers and influencers at customer companies.
## Authentication
Authenticate using OAuth 2.0 for secure API access.
```http
POST /services/oauth2/token
Content-Type: application/x-www-form-urlencoded
grant_type=password&
client_id=YOUR_CLIENT_ID&
client_secret=YOUR_CLIENT_SECRET&
username=YOUR_USERNAME&
password=YOUR_PASSWORD_AND_SECURITY_TOKEN
```
Append your security token to your password. Example: `MyPassword123` + `TOKEN456` = `MyPassword123TOKEN456`
## API Operations
POST to /sobjects/Lead/ or /sobjects/Opportunity/
PATCH to /sobjects/Lead/\{Id} with changed fields
GET with SOQL queries to retrieve records
Use Bulk API for large data sets (1000+ records)
## Security and Permissions
Control data access with profiles, permission sets, and sharing rules.
* **Standard Profiles:** Sales User, Marketing User
* **Custom Profiles:** Create industry-specific roles
* **Permission Sets:** Grant additional permissions
* **Permission Set Groups:** Bundle related permissions
* **Org-Wide Defaults:** Set base access levels
* **Role Hierarchy:** Grant upward access
* **Sharing Rules:** Criteria-based sharing
* **Manual Sharing:** Individual record access
## Tips and Tricks
**Data Quality:**
* Use validation rules to enforce data standards
* Set up duplicate matching rules
* Require key fields before saving
* Archive old records regularly
**Performance:**
* Use selective SOQL queries
* Index frequently queried fields
* Bulk API for large operations
* Monitor API call limits
**Automation:**
* Start with simple workflow rules
* Use Process Builder for complex logic
* Consider Flows for user interactions
* Test automations thoroughly
## Next Steps
Set up your Sales Cloud connection
Learn more about querying data
Explore sales automation scenarios
# Service Cloud
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
import { Tabs, Tab } from 'fumadocs-ui/components/tabs';
import { Accordion, Accordions } from 'fumadocs-ui/components/accordion';
# Service Cloud
Service Cloud is Salesforce's customer service platform. Deliver exceptional support across all channels with case management, knowledge base, and omni-channel routing.
>Service Cloud: Submit case via email/chat/web
Service Cloud->>Knowledge Base: Search for solution
alt Solution Found
Service Cloud-->>Customer: Auto-respond with article
else Needs Agent
Service Cloud->>Agent: Route to skilled agent
Agent->>Knowledge Base: Find relevant articles
Agent-->>Customer: Provide personalized support
Agent->>Service Cloud: Close case
end
`}
/>
## What You Can Do
Track and resolve customer issues across all channels
Create self-service articles and solutions
Route work intelligently based on skills and capacity
Monitor SLA compliance and team performance
## Case Management
Capture, track, and resolve customer service requests efficiently.
Ask MeshMesh: "Show me all high priority cases created today that haven't been assigned"
**Common Use Cases:**
* Multi-channel case capture
* Intelligent assignment and routing
* Automated escalation rules
* SLA tracking and enforcement
**Example Request:**
"Create a case for Sarah at Tech Corp about login issues with high priority"
## Knowledge Management
Build a centralized knowledge base for agents and customers.
Create articles with rich text, images, and videos. Use draft, review, and publish workflows with version control.
Deflect cases with proactive article suggestions. Customers find answers before contacting support.
AI recommends relevant articles during case resolution. Agents resolve issues faster with instant access to solutions.
## Omni-Channel Routing
Route customer requests to the right agent at the right time.
**Key Features:**
* Real-time presence and availability
* Skill-based routing
* Capacity-based assignment
* Priority queue management
**Example Request:**
"Route this case to an agent with billing expertise who's available now"
## Key Objects
**Case** represents customer service requests.
**Standard Fields:**
* CaseNumber (auto-generated)
* Subject, Priority, Status
* Origin (Email, Web, Phone, Chat)
* Type, Reason
**Example:** Track support tickets from initial contact to resolution.
**Knowledge\_\_kav** represents support documentation.
**Standard Fields:**
* Title, Summary, ArticleBody
* PublishStatus, ValidationStatus
* ArticleNumber (auto-generated)
* IsVisibleInPkb (public visibility)
**Example:** Store FAQ, troubleshooting guides, and how-to articles.
**Entitlement** defines service level agreements.
**Standard Fields:**
* Name, AccountId, StartDate, EndDate
* Type (number of cases or time-based)
* SlaProcess (milestones and actions)
**Example:** Track premium support agreements and response times.
## API Operations
MeshMesh handles API calls automatically. Describe what you need in plain English.
```json
POST /sobjects/Case/
{
"Subject": "Product not working",
"Status": "New",
"Priority": "High",
"Origin": "Web"
}
```
```http
PATCH /sobjects/Case/{Id}
{
"Status": "Closed",
"Resolution": "Issue resolved"
}
```
```http
GET /search/?q=FIND {password reset}
IN ALL FIELDS RETURNING
KnowledgeArticleVersion(Id,Title,Summary)
```
## Use Cases
Unified agent interface for email, chat, phone, and social media
Customer communities with knowledge articles and case tracking
Schedule technicians and track on-site service appointments
Monitor SLA compliance, resolution time, and agent productivity
## Tips and Tricks
**Case Management Best Practices:**
* Use assignment rules for automatic routing
* Set up escalation rules for SLA violations
* Create case templates for common issues
* Use macros for repetitive tasks
**Knowledge Base Optimization:**
* Use clear, descriptive article titles
* Include screenshots and videos
* Review and update articles regularly
* Track which articles resolve cases
**Omni-Channel Configuration:**
* Define realistic agent capacity
* Configure skills based on expertise
* Set priority for urgent work items
* Monitor queue performance
## Next Steps
Set up your Service Cloud connection
Learn more about case workflows
Explore customer service scenarios
# The Prompting Mindset
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
## Specificity Wins
The more specific your prompt, the better the result. Vague requests get generic responses. Specific requests get exactly what you need.
**Vague:**
```
Help me with Marketing Cloud
```
**Specific:**
```
Analyze my email performance for Q4 and identify which campaigns
had the lowest engagement — I want to improve them
```
**Vague:**
```
Set up Data Cloud
```
**Specific:**
```
Connect my Sales Cloud contacts and Marketing Cloud subscribers
into a unified profile in Data Cloud, with identity resolution
based on email address
```
## The Core Loop
Working with MeshMesh follows a simple pattern:
1. **Ask** — Describe what you want in plain language
2. **Review** — Look at what the agent created or proposed
3. **Refine** — Ask for changes, add context, or steer in a new direction
4. **Approve** — When it looks right, let it run
This is intentional iteration. Each round of feedback helps the agent understand exactly what you need.
## You Don't Need to Know How
You don't need to know the API, the field names, or the exact steps. Describe the outcome you want. The agent figures out the how.
```
I want to identify customers who bought last year but haven't
purchased anything in 2024. Then create a win-back campaign
targeting them with personalized offers based on their past purchases.
```
The agent will:
* Query your data to find these customers
* Analyze their purchase history
* Create the campaign with personalized content
* Ask you questions if anything is unclear
## Context Makes Everything Better
The agent can only work with what you give it. The more context, the better the output.
**Ways to add context:**
* **Drag and drop files** — Attach Word docs, Excel files, PDFs with requirements
* **Create references** — Save brand guides, process docs, and requirements for reuse
* **Tell the agent to read them** — "Read my brand guidelines before creating emails"
* **Build on the conversation** — The agent remembers everything you've discussed
```
I've attached our Q3 campaign results spreadsheet.
Analyze the data and tell me which segments performed best.
Then create a Q4 strategy that doubles down on what worked.
```
## The Agent Will Ask Questions
When something is unclear, the agent asks rather than assumes. This is a feature, not a bug.
If the agent asks:
* "Should I include historical data or just new records?"
* "Do you prefer approach A (simpler) or approach B (more flexible)?"
* "What's the VIP threshold for your customers?"
Answer directly. These questions help the agent deliver exactly what you need.
## Interrupt and Redirect Anytime
Changed your mind? Thought of something else? Just say so.
```
Actually, also segment this by region
```
```
Wait — before you deploy, let me review the Flow logic
```
```
That's not quite right. The VIP threshold should be $10,000, not $1,000
```
The agent adapts. You're steering the conversation, not executing a script.
## Next Steps
Learn the recommended workflow: research first, then implement
How to give the agent what it needs with references and files
See real prompts for common tasks
# Persona Feedback
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
Before finalizing content, you can get feedback from different perspectives. The agent can simulate how a specific persona would react to your content — a Gen Z consumer, a CMO, a technical user, or anyone else.
## How It Works
Tell the agent what persona you want feedback from and what content to review:
```
Get feedback on this email from a Gen Z consumer — is it authentic?
```
```
Have a CMO review this campaign for brand consistency
```
```
Get a UX designer's perspective on this landing page layout
```
The agent simulates that persona's perspective and provides feedback as if they were reviewing your content.
## Defining Effective Personas
The more specific the persona, the better the feedback. Include:
* **Role and experience** — What do they do? How long have they done it?
* **Values and priorities** — What do they care about?
* **Perspective** — How do they see the world?
* **Relevant context** — What makes them qualified to give this feedback?
**Good persona definitions:**
```
A CMO at a Fortune 500 company focused on brand consistency
and ROI. Skeptical of new marketing trends but open to
data-driven approaches. 15 years in the industry.
```
```
A Gen Z consumer (age 22) who values authenticity, sustainability,
and social responsibility. Skeptical of traditional marketing.
Prefers brands that align with their values.
```
**Avoid vague personas:**
* "A user" — Too generic
* "A designer" — What kind? What values?
* "Someone technical" — Be specific about their background
## Common Personas
### Marketing & Brand
```
Get feedback from a brand strategist focused on consistency
across channels
```
```
Have a performance marketer review this for conversion optimization
```
### Audience Segments
```
Review this from the perspective of a first-time buyer
who doesn't know our brand
```
```
How would a loyal customer of 5+ years react to this message?
```
### Technical & Stakeholders
```
Have an accessibility specialist review this for WCAG compliance
```
```
How would our VP of Sales react to this campaign?
```
## Using Feedback
Once you get persona feedback, use it to improve your content:
```
Based on the Gen Z feedback, make the tone more casual
and remove the corporate jargon
```
```
The CMO raised concerns about brand consistency. Update
the colors to match our brand kit.
```
## Multiple Perspectives
Get feedback from several personas to catch different issues:
```
Review this email from three perspectives:
1. A Gen Z consumer
2. A busy executive
3. An accessibility advocate
```
This helps you create content that works for your full audience, not just one segment.
Conflicting feedback between personas is valuable — it reveals trade-offs you need to consider. You might not be able to please everyone, but you can make an informed choice.
## Next Steps
Request visual outputs, reports, and more
See example prompts for campaigns and operations
# Plan Then Build
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
MeshMesh has two modes: **Plan** and **Build**. The recommended workflow is to start in Plan mode, get a solid plan approved, then switch to Build mode to execute.
## Why Plan First?
Complex Salesforce implementations have many moving parts. Rushing into build can lead to missing dependencies, overwriting existing configurations, or rework when you discover requirements mid-implementation.
Plan mode lets the agent explore your org, understand what exists, and design an approach that fits your reality.
## The Two Modes
**Plan mode** can read, explore, analyze, and propose — but cannot modify external systems. This is intentional. You review and approve before anything changes.
**Build mode** can deploy, create, modify, and execute. That's why you want a solid plan first.
Learn more about how agents work in Plan and Build modes in the [Agents concept guide](/docs/concepts/agents).
## The Workflow
1. **Start in Plan mode** — For complex implementations, choose Plan mode when creating a new task
2. **Review the plan** — Ask questions, request changes, or ask for clarification
3. **Approve and switch to Build** — When it looks good, say "let's build it"
You can switch back to Plan mode anytime if you need to rethink the approach.
## When to Use Plan Mode
**Use Plan mode for:**
* Multi-step implementations
* Anything that touches multiple systems
* Tasks where you're not sure what exists in the org
* Implementations with dependencies
**Skip Plan mode for:**
* Simple queries and analysis
* Quick content creation
* Straightforward tasks you know exactly how to do
## Next Steps
Give the agent what it needs with references and files
Leverage product-specific expertise
# Providing Context
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
The agent can only work with what you give it. Providing good context — requirements documents, brand guides, data files, examples — dramatically improves the quality of output.
## Quick Ways to Add Context
**Drag and drop files** — Attach Word docs, Excel files, PDFs, images directly into the chat.
```
I've attached the campaign brief. Create an email journey
that follows these requirements.
```
**Use references** — Save documents in MeshMesh for reuse across multiple tasks.
```
Read my brand guidelines before creating any emails
```
**Tell the agent what to focus on** — If you attach multiple files, be clear about priorities.
```
I've attached several documents. For this task, focus only
on the "Q4 Goals" document.
```
For detailed guidance on creating and managing references, see the [Using References guide](/docs/guides/using-references).
## What to Include
* **Requirements and specs** — Must-have features, acceptance criteria, constraints
* **Examples** — Sample emails, reports in your preferred format, templates to follow
* **Data** — Performance data to analyze, contact lists to segment
* **Brand and style guides** — Tone guidelines, terminology, formatting rules
* **Process documentation** — Approval workflows, naming conventions
## Best Practices
**Break large requirements into phases** — Don't try to do everything at once.
**Use accurate data** — The agent trusts what you give it.
**Keep images lightweight** — Large images slow things down.
## Next Steps
Create and manage reusable reference documents
Guide the agent with natural language
# Steering Conversations
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
You're in control of the conversation. Interrupt, redirect, and course-correct at any time. The agent adapts to your steering.
## Interrupt Anytime
You don't have to wait for the agent to finish. If you see it going in the wrong direction, stop it.
```
Wait — before you deploy, let me review the Flow logic
```
```
Hold on. I want to change the approach.
```
```
Stop. Let's think about this differently.
```
The agent pauses and waits for your guidance.
## Add Context Mid-Conversation
Realize you forgot to mention something? Add it.
```
Actually, also segment this by region
```
```
Oh, I should mention — we have an existing automation
that handles new leads. Don't duplicate that logic.
```
```
One more thing: the VIP threshold is $10,000 annual spend
```
The agent incorporates the new context and adjusts its work.
## Ask for Modifications
Not happy with what the agent created? Ask for changes.
```
That's not quite right. The VIP threshold should be $10,000,
not $1,000
```
```
Good start, but make the tone more casual and add emojis
```
```
This is too complex. Can you simplify it?
```
```
I like the structure but change the subject line to something
more urgent
```
The agent revises based on your feedback.
## Respond to Questions
The agent asks questions when it needs clarification. Answer directly.
**Agent:** "Should I include historical data or just new records?"
```
Just new records going forward. We'll handle historical
data in a separate phase.
```
**Agent:** "Do you prefer approach A (simpler) or approach B (more flexible)?"
```
Let's go with A for now. We can always make it more
flexible later if needed.
```
If you're not sure how to answer, say so. The agent can explain the trade-offs in more detail.
## Change Direction Entirely
Changed your mind about the approach? Just say so.
```
Actually, let's scrap this and try something different.
Instead of a Flow, let's use an Apex trigger.
```
```
I've been thinking about this wrong. Let's start over
with a focus on the mobile experience first.
```
The agent doesn't take it personally. It adapts and moves in the new direction.
## Build on Previous Work
The agent remembers the full conversation. Reference earlier work.
```
Use the same structure as the email we created earlier,
but for the mobile version
```
```
Apply the same segmentation logic to this new data extension
```
```
Remember the requirements from phase 1? Add those to this plan too.
```
## Ask for Explanations
If you don't understand something, ask.
```
Why did you choose this approach instead of using a formula field?
```
```
Can you explain what this Flow trigger does?
```
```
What would happen if we used the other option you mentioned?
```
The agent explains its reasoning and helps you make informed decisions.
## When to Start Fresh
Sometimes it's better to start a new conversation:
* The context has become cluttered with abandoned directions
* You want to try a completely different approach from scratch
* The conversation history is creating confusion
But most of the time, keep steering. The agent tracks context efficiently and can course-correct without starting over.
## Next Steps
Get feedback from simulated personas before finalizing
See example prompts for analysis and audit tasks
# Version 0.19.10 - Connection Fix
# Version 0.19.10
**December 6, 2025**
This patch resolves connection issues that could occur when restarting the agent environment. In some cases, the agent wouldn't properly reconnect after a restart, leaving tasks unable to execute. This is now fixed.
Also includes a fix for handling multiple Salesforce orgs. If you have several orgs connected, they're now fetched and managed correctly from the desktop client.
## What's Fixed
* **Agent Restart** - Fixed connection issues when restarting the agent environment
* **Multiple Orgs** - Fixed handling when multiple Salesforce orgs are configured
# Version 0.19.11 - Onboarding Improvements
# Version 0.19.11
**December 9, 2025**
The onboarding experience now includes brand kit creation. During your first run, you can set up your brand colors, fonts, logos, and guidelines so the AI generates on-brand content from the start. No need to configure this later - get it right during setup.
You can also name your workspace during onboarding. This is helpful if you're setting up workspaces for different teams or projects and want to keep them organized from day one.
Brand kit previews have been improved to give you a better sense of how your settings will look before you save them. And we've fixed an issue on macOS where closing the app window didn't fully quit the application - it now exits properly.
## What's New
* **Brand Kit Onboarding** - Create your brand kit during the first run experience
* **Workspace Naming** - Name your workspace during setup
* **Better Brand Previews** - Improved preview of brand kit settings before saving
## What's Fixed
* **App Quit** - Closing the window now properly quits the app on macOS
## Security
* General security improvements and CI/CD hardening
# Version 0.19.13 - Live Task Context
# Version 0.19.13
**December 10, 2025**
You can now change connections and references while a task is actively running. Previously, modifying task context required stopping the task first. Now you can add a new reference, remove one that's no longer needed, or swap to a different connection without interrupting the AI's work. The agent picks up context changes on the fly.
Reference loading is now significantly faster thanks to smart caching. References are cached locally and only re-fetched when they've actually changed. For workspaces with large document sets, this means much faster task startup times.
This release also includes a critical fix for Salesforce connections. Tokens that expired during periods of inactivity are now refreshed automatically, so connections don't silently fail after you've been away.
## What's New
* **Live Task Context** - Add or remove connections and references while tasks run
* **Reference Caching** - Faster reference loading with local caching and smart invalidation
## What's Fixed
* **Salesforce Tokens** - Fixed automatic token refresh after periods of inactivity
# Version 0.19.17 - Planning Mode
# Version 0.19.17
**December 11, 2025**
Introducing Planning Mode. Before executing a task, the AI now outlines its approach first. You'll see what it intends to do, which tools it plans to use, and how it will tackle the problem. This gives you a chance to course-correct before any work begins - catch misunderstandings early and guide the AI toward better results.
Visual transitions make it clear when the AI is in planning mode versus execution mode. The switch is smooth and intuitive, so you always know what state you're in.
This release also adds a Beta Release Channel. If you want early access to new features before they hit the stable release, you can opt into beta builds from settings. And for security, we've moved authentication storage to your system's native keychain using SafeStorage encryption - credentials are now protected by OS-level security instead of local file encryption.
## What's New
* **Planning Mode** - AI plans before executing with clear visual state transitions
* **Beta Channel** - Opt into beta releases for early access to new features
* **Keychain Storage** - Auth credentials now use system keychain instead of local storage
## What's Fixed
* **Salesforce Connections** - Fixed handling of invalidated connections and CLI compatibility issues
# Version 0.19.18 - File Upload Fix
# Version 0.19.18
**December 11, 2025**
This patch fixes file upload functionality with the agent. Documents and assets now upload correctly during AI tasks.
## What's Fixed
* **File Uploads** - Resolved upload issues with the agent environment
# Version 0.19.19 - Agent Coding & Salesforce Scopes
# Version 0.19.19
**December 11, 2025**
The AI agent can now write and execute code as part of your tasks. Need to transform data, generate a script, or analyze something programmatically? The agent can write code, run it, and use the results to complete your task. This is especially useful for technical workflows where you need more than just information retrieval.
For Salesforce users connecting via external connected apps, you can now configure custom OAuth scopes. This gives you control over exactly which permissions your connection has - useful when you need access to CDP or Ingest APIs, or when you want to limit scope for security reasons.
## What's New
* **Agent Coding** - AI can write and execute code during task execution
* **Salesforce Custom Scopes** - Configure specific OAuth scopes for CDP, Ingest API, and other permissions
## What's Fixed
* **Connection Steps** - Resolved build issues affecting connection configuration
# Version 0.19.21 - Plan Mode & Document Support
# Version 0.19.21
**December 12, 2025**
Planning Mode gets a visual upgrade with smoother transitions between planning and execution states. The mode switch is now more intuitive, making it clearer when the AI is thinking through its approach versus actively making changes.
You can now upload PDF and CSV files directly into tasks. Drop a report, spreadsheet, or data export into your task and the AI will read and analyze it alongside your other references. This is useful for tasks that involve reviewing documents, analyzing data, or working with files from external systems.
We've also improved stability around message streaming. When you stop a task mid-stream, pending messages are now handled more gracefully. The offline connectivity banner has been refined to give you clearer status information, and several file reference bugs have been squashed.
## What's New
* **Plan Mode Improvements** - Smoother visual mode switching and refined planning instructions
* **PDF & CSV Support** - Upload documents directly into tasks for AI analysis
* **Offline Banner** - Clearer connectivity status display
## What's Fixed
* **Message Streaming** - Better handling when tasks are stopped mid-stream
* **File References** - Fixed deletion and duplicate insert issues
* **Security Updates** - Dependabot patches applied
# Version 0.19.22 - Interactive Questions & Default Plan Mode
# Version 0.19.22
**December 25, 2025**
This release introduces interactive questions, allowing the AI to ask you for clarification when it needs more information during a task. Instead of making assumptions, the agent can now pause and ask specific questions to ensure it understands exactly what you need.
Plan mode is now enabled by default for all tasks. When you start a new task, the AI will first create a plan outlining what it intends to do, giving you the opportunity to review and approve the approach before any changes are made. This leads to more predictable and accurate results.
We've also improved the planning experience with better artifact display and fixed several issues around updating plans mid-task.
## What's New
* **Interactive Questions** - AI can now ask for clarification during tasks instead of making assumptions
* **Plan Mode by Default** - All tasks now start in planning mode for better accuracy
* **Improved Planner** - Smarter planning tools for more reliable task execution
## What's Fixed
* **Artifact Display** - Fixed display issues when viewing artifacts in plan mode
* **Plan Updates** - Resolved issue with updating plans mid-task
* **Large Payloads** - Fixed errors when processing large amounts of data
* **Proxy Timeouts** - Improved stability for long-running operations
# Version 0.19.28 - Better Reference Management
# Version 0.19.28
**January 7, 2026**
Managing references just got easier. We've redesigned the reference selection flow to help you find and add the right documents to your tasks faster. You can also now edit references directly without leaving your current context.
Light mode users will notice improved styling throughout the app, and we've fixed several issues with uploading new references.
## What's New
* **Better Reference Selection** - Redesigned flow makes it easier to find and add references to tasks
* **Direct Edit** - Edit references without leaving your current context
## What's Fixed
* Light mode styling issues
* Reference upload reliability
# Version 0.19.30 - Intel Mac Support & Auto-Updates
# Version 0.19.30
**January 8, 2026**
MeshMesh Studio is now available for Intel-based Macs. Whether you're on Apple Silicon or an older Intel machine, you can download the right version from our download page and get started immediately.
We've also added automatic background updates. When you quit the app, it will download any available updates so you're always on the latest version the next time you open it. No more manual update prompts interrupting your workflow.
Workspace switching has been improved for teams managing multiple workspaces, and we've squashed several bugs around plan mode transitions.
## What's New
* **Intel Mac Support** - Desktop app now runs natively on Intel-based Macs
* **Auto-Update on Quit** - Updates download automatically in the background when you close the app
* **Workspace Switching** - Smoother experience when switching between workspaces
## What's Fixed
* Plan mode transitions to build mode
* Account switching between workspaces
* Connection retry handling
# Version 0.19.31 - Simplified Onboarding
# Version 0.19.31
**January 10, 2026**
We've simplified the onboarding experience to get you up and running faster. Connections are now auto-selected by default, and Salesforce is no longer required during setup. You can always add or configure connections later from your workspace settings.
This release also fixes an issue with connecting to Salesforce on Mac.
## What's New
* **Simplified Onboarding** - All connections are now auto-selected by default during setup
* **Optional Salesforce** - Salesforce connection is no longer mandatory to complete onboarding
## What's Fixed
* Issue with connecting to Salesforce on Mac
* Route handling issue
# Version 0.19.6 - Reference Sharing
# Version 0.19.6
**December 4, 2025**
You can now share references with team members in your workspace. When you've created a useful reference document - brand guidelines, technical specs, standard operating procedures - you can share it so others can use it in their tasks too.
Shared references appear in team members' reference lists and can be attached to tasks just like personal references. This makes it easier to maintain consistent context across your team's work.
## What's New
* **Reference Sharing** - Share references with team members in your workspace
# Version 0.19.7 - Task Subagents
# Version 0.19.7
**December 4, 2025**
The AI can now spawn sub-tasks when it needs to explore a topic more deeply. For complex questions that require research across multiple areas, the agent can break the work into focused subtasks, execute them in parallel, and combine the results into a comprehensive answer.
This is especially useful for exploratory work where the AI needs to investigate several angles before synthesizing a response. Instead of a shallow pass over everything, you get deeper analysis of each component.
## What's New
* **Task Subagents** - AI can spawn sub-tasks for deeper exploration and research
## What's Fixed
* **Actions Display** - Fixed actions not displaying after page refresh
# Version 0.19.8 - File Browser
# Version 0.19.8
**December 5, 2025**
You can now browse and view files that the AI agent creates during tasks. When the agent generates reports, exports data, or creates any files as part of its work, you can see them in the new file browser. No more digging through logs to find outputs - everything the agent creates is visible and accessible.
The file browser only appears for tasks where the agent has file creation capabilities, so it stays out of the way when it's not relevant.
This release also fixes issues with the agent environment not restarting correctly, and resolves problems when importing custom MCP servers.
## What's New
* **File Browser** - Browse and view files created by the AI agent during tasks
## What's Fixed
* **Agent Restart** - Fixed issues with the agent environment not restarting correctly
* **Custom MCP Import** - Resolved issues when importing custom MCP server configurations
# Version 0.20.0 - Blueprints & New Experience
# Version 0.20.0
**January 27, 2026**
This is our biggest release yet. Blueprints bring end-to-end implementation guides that take you from planning through execution - whether you're building a Data 360 implementation, deploying Agentforce agents, or running a Marketing Cloud campaign.
We've also revamped the entire interface for a cleaner, more focused experience. New users can now get started instantly with demo connections - no need to connect your production org to try things out.
## What's New
### Blueprints
Blueprints are end-to-end implementation and operation execution guides that help you accomplish real outcomes:
* **Plan & Build a Data 360 Implementation** - Unified customer data across your Salesforce ecosystem
* **Create & Deploy Agentforce Agents** - Build AI agents for service, sales, or custom use cases
* **Plan & Build Marketing Cloud Campaigns** - From journey design to execution
* **Run Tech Debt Audits** - Analyze and clean up your Salesforce org
* **Analyze Sales Pipelines** - Get insights from your opportunity data
* **Build Loyalty Programs** - Design and implement customer loyalty initiatives
### Revamped Interface
* **Cleaner Dashboard** - Task list sidebar shows your work prioritized by status
* **Context Cards** - Quick access to Brand Kit, Business Context, and Integrations
* **Improved Task View** - Two-column layout with references on the left and focused chat on the right
* **Better Mode Switching** - Clear Plan and Build modes with descriptions
### Skills for Salesforce Clouds
New specialized skills that understand how to work with each Salesforce product:
* **Agentforce** - Build and deploy AI agents with custom actions directly to your org
* **Marketing Cloud Next** - Work with segments, bulk contact ingestion, and identity resolution
* **Marketing Cloud Engagement** - Manage journeys, automations, triggered sends, and campaigns
* **Data Cloud** - Configure activations, data connections, and calculated insights
* **Metadata Deployment** - Deploy Apex classes, Flows, Lightning components, and Permission Sets
### Document Skills
Create professional documents without leaving the platform:
* **Word Documents** - Create, edit, and analyze DOCX files with tracked changes support
* **Excel Spreadsheets** - Build spreadsheets with formulas, formatting, and data analysis
* **PowerPoint Presentations** - Create presentations from scratch or using templates
* **PDF Export** - Export artifacts as PDF documents
### Enterprise Features
* **BYOK (Bring Your Own Keys)** - Use your own API keys for AI providers with workspace-level controls
* **Demo Connections** - New users get instant access to Salesforce and Marketing Cloud demos
* **Public References** - Share reference documents across your organization with preview images
### Other Improvements
* **Claude Opus 4.5** is now the default AI model for improved performance
* Simplified team invitations
* Renamed "Chats" to "Agents" for clarity
* Cleaner header and navigation throughout
## What's Fixed
* AI no longer gets stuck asking the same question repeatedly
* Improved Salesforce CLI connection reliability
* Better task auto-naming from your prompts
* Various interface improvements and polish
# Version 0.20.4 - Sandbox Support
# Version 0.20.4
**February 3, 2026**
This release adds support for Salesforce sandbox environments - perfect for testing before going to production.
## What's New
### Salesforce Sandbox Support
Connect to your Salesforce sandbox environments directly from the desktop app. Test your implementations safely before deploying to production.
### Bulk Task Management
Select and delete multiple tasks at once. Keep your workspace clean with easier task management.
## What's Improved
* Smoother agent conversations with better message handling
* More reliable connection handling
* Better handling of file uploads
# Version 0.20.9 - Browser Use & GitHub Integration
# Version 0.20.9
**February 19, 2026**
Agents just got a lot more capable. With Browser Use, your agent can now open Salesforce in a real browser — navigating Setup, clicking through Lightning pages, filling forms, and taking screenshots — all while you watch in real-time. Plus, connect your private GitHub repos and learn faster with video tutorials built right into the app.
## What's New
### Browser Use
Your agent can now interact with Salesforce directly in the browser. Navigate Setup menus, configure settings, inspect pages, and more — with a live browser view so you can watch every step.
* Open any Salesforce page and interact with it in real-time
* Watch your agent work with a live browser stream
* Agent takes screenshots and reads page content to make decisions
* Works with any connected Salesforce org
### GitHub Integration
Connect your private GitHub repositories to your workspace. Your agent can access code, configurations, and documentation stored in your repos — perfect for working with custom Apex classes, Lightning Web Components, or DevOps configurations.
### Video Tutorials
Contextual tutorial videos are now embedded throughout the app — on the dashboard, in settings, during onboarding, and across the documentation. Learn how to get the most out of MeshMesh without leaving the app.
## What's Improved
* Restyled login experience with a fresh, modern look
* Smarter plan mode with improved guardrails
* More reliable sandbox provisioning
* Upgraded AI engine for better agent performance
* Long URLs in chat messages now display correctly
## What's Fixed
* Fixed Salesforce credential requests occasionally timing out
* Fixed chat sessions becoming unresponsive after certain errors
* Fixed multi-file upload issues
* Fixed video playback in documentation
* Improved overall agent stability and connection reliability
# February 20, 2026 - HTML Exports & Stability
# February 20, 2026
## What's Improved
* HTML artifacts now download as `.html` files instead of PDF for better compatibility and usability
## What's Fixed
* Fixed artifact display and conversion issues
* Improved agent stability and reduced memory-related crashes
# February 23, 2026 - Large PDF Handling
# February 23, 2026
## What's Improved
* Better handling of large PDF uploads — oversized files are now handled gracefully with clear feedback
## General
* Performance and reliability improvements
# February 24, 2026 - Slack Support & Security
# February 24, 2026
## What's New
* **Built-in Slack support** — get help directly from the app through a dedicated Slack support channel
## What's Fixed
* Chat input no longer carries over text between different tasks
* Security improvements
# March 4, 2026 - Selectable Artifacts & Long Session Memory
# March 4, 2026
## What's New
* **Selectable artifact text** — select and copy text directly from artifacts and references
* **Long session artifact recall** — agents can now reference and recall earlier artifacts in extended sessions
## What's Improved
* Smarter chat scrolling that respects your scroll position — no more jumping to the bottom while you're reading
* Individual demo environments — each user gets their own demo connection for a better trial experience
## General
* Security improvements
# March 5, 2026 - Screenshot Sharing, File Previews & Smarter Agents
# March 5, 2026
## What's New
* **Inline screenshot sharing** — share screenshots directly in your task chat for visual context
* **Smarter agent evaluation** — agents now evaluate their own work with structured checkpoints for more reliable results
## What's Fixed
* File browser now shows image previews correctly
* Dark mode PDF exports render with correct background colors and page breaks
* PDF downloads now export the actual document instead of a screenshot
* Deleted tasks are now immediately removed from the sidebar
* Fixed BYOK (Bring Your Own Key) "Fetch Failed" error in the web console
* Fixed RSS feed errors on the release notes page
# March 10, 2026 - Cross-Agent Search
# March 10, 2026
## What's New
* **Cross-agent keyword search** — search across all your agents and agent messages instantly to find exactly what you need
> **Pro tip:** Use **Cmd+K** (Ctrl+K on Windows) to quickly search agents and messages, switch between tasks, and jump to references or blueprints from anywhere in the app.
## What's Improved
* More accurate usage tracking and cost reporting
# March 11, 2026 - Chat Navigation & Task List Pagination
# March 11, 2026
## What's New
* **Prompt navigation** — use Cmd/Ctrl+Up/Down to quickly jump between your messages in a task chat
* **Task list pagination** — browse older tasks with "Load more" in the sidebar
## What's Fixed
* Demo connections are now protected from accidental edits or deletion
* Improved stability when sandbox environments encounter errors
* Fixed tasks not loading correctly when opened from search
# March 12, 2026 - Connection Reliability & Security Hardening
# March 12, 2026
## What's Improved
* **More reliable Salesforce connections** — connections now properly await provisioning before allowing agent interaction, preventing timeouts and unresponsive sessions
* **Better sandbox stability** — sandbox environments now handle errors gracefully instead of triggering repeated authentication failures
## General
* Security improvements including enhanced input validation, stricter authorization checks, and hardened integration configurations
# March 13, 2026 - Agent Stability & Reliability
# March 13, 2026
## What's Improved
* **More stable long-running tasks** — agents can now handle extended conversations without slowing down or crashing, with significantly lower token costs
* **Better agent reliability** — agents run with increased resources and smarter error recovery, reducing interruptions during complex tasks like browser automation
* **Reliable streaming** — long-running agent responses no longer time out, with clearer error messages when tasks are interrupted
## What's Fixed
* **File downloads now work reliably** — downloading assets like documents, spreadsheets, and PDFs from agent responses no longer fails
# March 16, 2026 - Agent Rename, File Types & Auto-Recovery
# March 16, 2026
Your agents just got easier to organize and more resilient. Rename agents right from the sidebar, upload more file types as references, and if anything goes wrong with your agent's environment, it recovers automatically — no action needed.
## What's New
### Agent Rename
Rename any agent directly from the sidebar — right-click or use the dropdown menu. No more navigating into an agent just to change its name.
### More Reference File Types
Upload JSON, YAML, HTML, and audio files (MP3, WAV) as references alongside your existing documents and images. Great for sharing configuration files, API specs, or audio briefs with your agent.
### Automatic Sandbox Recovery
If your agent's cloud environment goes down unexpectedly, MeshMesh now detects the failure and automatically restores from backup — your agent picks up right where it left off without any manual steps.
## What's Improved
* **Faster, more reliable search** — searching with long queries no longer fails, with smarter debouncing for smoother results
* **Salesforce setup guide** — new step-by-step documentation for connecting your Salesforce org with OAuth and PKCE
## What's Fixed
* Security patches across multiple dependencies
# March 23, 2026 - Reference Sharing, @ Mentions & Agent Images
# March 23, 2026
This release is all about making your best work reusable and shareable. Turn any agent run into a recipe your team can follow, share references with Google Drive-style permissions, mention them inline with @, and preview brand kits visually. Plus, your agent can now upload and share images directly in the conversation.
## What's New
### Recipe Creator
Turn any completed agent run into a reusable recipe. After your agent finishes a task you're happy with, just say "turn this into a recipe" — the agent reviews what it did, asks whether you want exact steps or a generalized template, and saves a structured guide with prerequisites, step-by-step instructions, verification checks, and troubleshooting tips. Share recipes with your team so anyone can repeat the same outcome without starting from scratch.
### Reference Sharing
Share any reference with your team using familiar Google Drive-style permissions. Set access to private, organization-wide, or share with specific team members — each with view or edit permissions. Copy a shareable link to send to anyone in your org.
### Stars & Favorites
Click the star icon on any reference to mark it as a favorite. Your starred references appear in a dedicated view so you can quickly find the ones you use most.
### Tags & Filtering
Organize references with color-coded tags — Salesforce products like Sales Cloud, Marketing Cloud, and Data Cloud get their own colors, and you can create custom tags too. A new sidebar lets you filter by type, tag, or both to find exactly what you need.
### @ Mention References in Chat
Type `@` in the chat input to search and insert any reference directly into your prompt. References already attached to your task appear first, and adding an external reference is just one click away. Your agent gets the full reference content as context.
### Agent Image Sharing
Your agent can now upload and use images during tasks. Images appear inline with a download button on hover, so you can save them directly to your computer.
### Brand Kit Previews
Brand kits now show live HTML previews that auto-cycle through up to 5 examples. Hover across the preview to jump between examples, or let them rotate automatically. See exactly how your brand looks before adding it to a task.
## What's Improved
* **Fullscreen artifact preview** — expand any HTML artifact to full screen for a closer look
* **Faster reference browsing** — server-side search and pagination for large reference libraries
* **New button in sidebar** — quickly create references, brand kits, or business context without leaving the modal
* **Right-click context menus** — star, share, or delete references directly from the context menu
* **Smarter recipes** — recipe references now show polished previews with auto-generated images
## What's Fixed
* More reliable sandbox recovery when resuming interrupted sessions
* Improved stability for long-running agent environments
# March 24, 2026 - Archive Agents, Success Celebrations & What's New
# March 24, 2026
Tidy up your workspace by archiving agents you're not using, get a satisfying celebration when your agent finishes a task, and never miss an update with the new "What's New" modal on launch.
## What's New
### Archive Agents
Hide agents you're not actively using without deleting them. Right-click any agent in the sidebar to archive it — archived agents disappear from your active list but remain fully searchable via Cmd+K cross-agent search. Unarchive anytime from the collapsible "Archived" section at the bottom of the sidebar. Bulk-select multiple agents to archive them all at once.
### Success Celebrations
When your agent completes a task successfully, you'll see a celebration with a summary of what was accomplished — including key highlights and a checklist of what was done. Rate your agent's work with a quick star rating and optional feedback to help improve future results.
### What's New on Launch
A new "Since You've Been Gone" modal greets you on launch with a summary of updates you missed since your last visit. Browse through release highlights with images and descriptions, and download the latest version directly from the modal. You can turn this off in Settings if you prefer.
### Browser Activity Indicator
When your agent is using the browser and you have a reference or artifact overlay open, a floating indicator appears at the top of your screen showing a live preview of the browser session. Click it to jump straight to the browser view. The indicator shows whether the agent is actively browsing or idle.
### Quick-Add References
Your recent and starred references now appear as quick-access pills when adding references to a task — just click to add instantly without searching.
## What's Improved
* **Clearer integrations page** — built-in integrations (Salesforce, Marketing Cloud, GitHub) are now separated from custom user-created connections, each with dedicated icons and descriptions
* **Skills always on** — all agent skills are now enabled by default, no feature flag or settings toggle needed
* **Fewer credential prompts** — smarter detection means the credential modal only appears when truly needed, with a skip button as an escape hatch
* **Update notifications** — the "New version available" toast is back with an install-and-restart button so you always know when an update is ready
## What's Fixed
* More reliable update notifications when a new version is available
* Improved stability for long-running browser sessions with overlays
# March 26, 2026 - B2C Commerce, Auto Storage Resize & Live Progress
# March 26, 2026
Connect to Salesforce B2C Commerce Cloud, let your agent's storage grow automatically, and see exactly what your agent is doing in real time.
## What's New
### B2C Commerce Cloud Integration (Beta)
Your agent can now work directly with Salesforce B2C Commerce Cloud. Connect your Commerce Cloud instance and your agent can deploy code versions, manage cartridges, configure shopper login, spin up on-demand sandboxes, run jobs, and deploy storefronts. Credentials are handled automatically — just add your connection and start working.
### Automatic Storage Upgrades
Your agent's storage now automatically upgrades when disk usage gets high. The agent pauses at a safe point, the storage resizes, and work resumes — no manual intervention needed. You'll see a toast notification when a resize is in progress and when it completes.
### Live Progress During Agent Work
When your agent is writing files or running commands, you'll now see real-time progress in the loading indicator — like how long the operation has been running and how many lines have been written. Progress appears after 5 seconds to avoid flickering on quick operations.
### Faster, Uninterrupted Streaming
Your agent can now handle longer, more complex tasks without interruption — even multi-minute responses stream smoothly from start to finish.
## What's Improved
* **Better failover** — automatic fallback to alternative AI providers if any provider is overloaded, so your agent keeps working without interruption
* **More reliable sign-up** — faster account creation for new users and org invitations, no duplicate workspaces or Slack channels
* **Search & navigation docs** — new guide covering Cmd+K command palette and dashboard search features
# March 27, 2026 - Smart Action Escalation & B2C Commerce Improvements
# March 27, 2026
Your agent just got smarter about keeping you in the loop. Critical actions are now automatically detected and escalated for your confirmation before anything happens.
## What's New
### Smart Action Escalation
Your agent now automatically identifies critical actions — like changes that could affect your live environment — and pauses to request your confirmation before proceeding. Think of it as an extra safety harness: even if a task involves a sensitive operation, your agent will flag it and wait for your go-ahead. You stay in control without having to watch every step.
## What's Improved
* **More reliable B2C Commerce connections** — B2C Commerce Cloud connections are now more stable and secure
* **Longer connection names** — Salesforce connection names now support more characters, so you can use more descriptive names for your orgs
* Reliability and security improvements across the platform
# March 29, 2026 - Slack Integration
# March 29, 2026
Bring your own Slack apps and bots, and let MeshMesh orchestrate them.
## What's New
### Slack Integration
Connect your existing Slack apps and bots to MeshMesh and let your agent orchestrate them. Send messages, manage channels, browse canvases, kick off workflows, and reply to threads — all without leaving MeshMesh. Bring any Slack app or bot you already use and your agent will work with it directly.
## What's Improved
* **Longer sessions** — stay signed in longer without unexpected logouts
* **More reliable connections** — connection setup is now more consistent and dependable
* Reliability improvements across the platform
# March 31, 2026 - Reference List View & Search
# March 31, 2026
Find exactly the reference you need — faster than ever.
## What's New
### Reference List View & Search
You can now switch between grid and list views when browsing your references and agent outputs. The new list view gives you a clean, scannable table with thumbnails, names, types, and timestamps — perfect for when you have a lot of references to sort through. Plus, a new search bar lets you find any reference instantly by name, description, or content.
### Editable Time Saved
After your agent wraps up a task, you can now adjust the time-saved estimate to better reflect your experience. You can also leave a quick star rating, pick feedback tags, and add notes — all of which help us make your agent even better over time.
## What's Improved
* **More stable agent sessions** — agent environments recover more reliably if anything goes wrong mid-session
* Stability and reliability improvements across the platform
# April 1, 2026 - Better Connection Error Handling & More Resilient Agents
# April 1, 2026
## What's New
### Better Connection Error Handling
Integrations now automatically detect common corporate proxy and VPN setups (Zscaler, Netskope, etc.) and handle authentication and SSL configuration without manual steps. When errors do occur, you'll see clear messages explaining what went wrong and how to fix it.
### More Resilient Agents
Agents now stop themselves after repeated denied actions instead of retrying in a loop. If the connection to your agent is lost mid-task, the task is paused rather than marked as failed — send a message to pick up where you left off. Any pending confirmation dialogs are also automatically dismissed so nothing is left hanging.
## What's Improved
* Stability and performance improvements
# April 3, 2026 - Performance & Reliability
# April 3, 2026
## What's Improved
* **Fewer interruptions** — agent sessions are more stable, resolving occasional "Failed to send" errors some users experienced
* **Faster response times** — agents respond more quickly to your messages
* **Smoother reconnections** — brief network changes are handled automatically without disrupting your work
* **Additional guardrails around browser use** — agents have extra checks when browsing to keep your sessions running smoothly
# April 8, 2026 - Slack Enterprise Grid & Editor Improvements
# April 8, 2026
## What's New
### Slack Enterprise Grid Support
If your organization uses Slack Enterprise Grid, MeshMesh now automatically detects and routes messages to the correct workspace within your grid. No extra configuration needed — connect your Slack app the same way you always do and MeshMesh handles workspace routing behind the scenes.
This works alongside existing standard Slack workspace connections, which continue to work as before.
### Search & Find-Replace in Editors
Code and markdown editors now support full search (Cmd/Ctrl+F) and find-replace, making it much easier to work with larger reference documents.
* **Find** — jump between matches with next/previous
* **Replace** — replace individual matches or all at once
* Works in both the code editor and markdown editor
The reference editing experience has also been unified — the edit button now opens a consistent View/Edit modal across all reference types.
## What's Improved
* **Faster task recovery** — interrupted agent sessions resume more quickly when you send a follow-up message
* **More consistent backups** — agent work is saved more frequently, reducing the chance of lost progress
* Stability and performance improvements
# April 15, 2026 - Business Intelligence & Enhanced File Browser
# April 15, 2026
## What's New
### Business Intelligence
A new intelligence dashboard that gives you deep visibility into your Salesforce org — powered by your agent:
* **Agent-powered analysis** — your agent connects to your org, scans metadata, and generates rich visual reports automatically
* **Pre-built templates** — get started instantly with Org Health, Permissions Audit, Field Utilization, Automation Inventory, Code Quality, and Data Model analysis
* **Custom analysis** — describe what you want to understand and the agent builds a tailored dashboard
* **Rich visualizations** — metric cards with trend sparklines, bar charts, donut charts, line charts, heatmaps, sortable tables, and narrative analysis
* **Actionable recommendations** — each report includes prioritized recommendations with impact and effort scores, and one-click buttons to have your agent fix the issues
* **Health scoring** — a 0–100 health score for your org that updates with each scan
* **Auto-refresh** — schedule recurring scans to continuously monitor your org
### Enhanced File Browser
The file browser has been significantly upgraded with deeper integration into the task experience:
* **Integrated view** — browse workspace files alongside the chat without leaving the task, with a smooth animated toggle between reference and file views
* **Drag-and-drop uploads** — drop files directly onto folders in the file tree, with support for ZIP extraction and automatic security scanning
* **File management** — move files between folders by dragging, delete with a confirmation dialog, and drag-and-drop reordering
* **Auto-refresh** — the file tree updates every 5 seconds, and new files appear with a smooth fade-in animation
* **Live agent edits** — see file changes in real time as the agent works, with green gutter markers highlighting changed lines
* **Resizable panels** — drag to resize the file tree, editor, and chat to fit your workflow
* **Tabbed editing** — open multiple files at once and switch between them
* **Command palette** — press Cmd+P to quickly find and open any file
* **Rich file previews** — view images, SVGs, CSVs as formatted tables, and markdown with a live split preview
### Drag-and-Drop on Dashboard
You can now drop images and files directly into the prompt area when starting a new task. Images appear as inline thumbnails, and other files are automatically uploaded as references.
### Improved B2C Commerce Cloud Connections
B2C Commerce Cloud connections are now more reliable, with clearer error messages when credentials need attention and automatic token refresh to keep sessions running smoothly.
## What's Improved
* **Smoother file editing** — saved changes now appear immediately when you reopen a file
* **Cleaner error messages** — a single notification on save errors instead of repeated alerts
* Security and dependency updates
# April 21, 2026 - Introducing Cloud Control 5 (Preview)
# April 21, 2026
## What's New
### Cloud Control 5 (Preview)
We're introducing **Cloud Control 5** — a significant step up from Cloud Control 4 (CC4) in how much of Salesforce your agent can confidently work across. It's available today as an opt-in preview, so you can try it on specific agents while everything else continues to run on the stable CC4.
**What's better with Cloud Control 5**
* **Commerce** — stronger results on B2B and B2C Commerce Cloud work
* **Service & Field Operations** — much improved handling of Service Cloud and Field Service Lightning scenarios
* **Industry Clouds** — meaningful new coverage for Manufacturing, Financial Services, Experience Cloud, and Nonprofit (NPSP)
* **Data Cloud, in depth** — noticeably better at specialised Data Cloud work: identity resolution, segmentation, calculated insights, Einstein Studio, and personalization
* **Marketing Cloud, in depth** — deeper support for Journey Builder, Automation Studio, Campaign Brief Builder, Script Activities, Package Manager, and Marketing Cloud Next
* **Agentforce development** — stronger at building and deploying custom Agentforce agents, including Agent Script
* **Account Engagement (Pardot)** — broader support including advanced Engagement Studio extensions
* **Einstein AI** — clearer handling of end-to-end Einstein setup and usage
* **Salesforce Knowledge** — better at authoring, organising, and maintaining Knowledge content
* **MuleSoft, end to end** — substantially stronger across the MuleSoft platform: API management, CI/CD, Composer, DataWeave, IDP, monitoring, RPA, and runtime
* **Platform power-users** — new strength on advanced patterns like Decision Tables and platform data investigations
* **Stronger recall on the long tail** — Cloud Control 5 draws on a much larger pool of historical Salesforce know-how, so it's more likely to have a useful answer for niche or older scenarios
**For most Salesforce work today, we recommend giving Cloud Control 5 a try** — especially if your workflows touch Agentforce (including Agent Script), Commerce, Service, Data Cloud, Marketing Cloud, or MuleSoft.
### How to try it
On the dashboard, open the Cloud Control selector next to the Build / Plan button and pick **CC5 (Preview)** before you start your next agent. Your choice is fixed for the lifetime of that agent — to switch, simply create a new one. The active Cloud Control version stays visible as a small badge in the agent's chat toolbar.
### A note on preview
Cloud Control 5 is labelled Preview for a reason: it's larger and more current, but still settling in. If anything feels off on a particular task, switch back to CC4 for that one and let us know what you saw — your feedback directly shapes what we roll out next.
## What's Improved
* **Always know what your agent knows** — a small `CC4` / `CC5` badge on every agent makes the current Cloud Control version unmistakable
* **Safe defaults** — new agents continue to use the stable Cloud Control 4 unless you explicitly opt in, so existing workflows are unaffected
* **Precise file drops in file view** — dropping a file onto a folder uploads into that folder, onto a file uploads into its parent, and onto empty space uploads to the workspace root. The reference uploader no longer grabs drops meant for the file browser.
* **Uninterrupted B2C Commerce Cloud sessions** — long-running Commerce Cloud agents previously stalled every \~30 minutes when authentication expired. Now your connection is refreshed automatically before expiry, so extended tasks keep running without manual re-auth or silent failures.
# April 27, 2026 - New Public App integration for Marketing Cloud
# April 27, 2026
## What's New
### Marketing Cloud now uses a Public App integration
Marketing Cloud connections in MeshMesh now use a **Public App** integration with personal sign-in, replacing the previous shared-credentials setup. Each user authenticates with their own Marketing Cloud account, so every action your agent takes in Marketing Cloud is traceable to a real user.
**Why we changed this**
* **No client secret** — the Public App integration uses the Authorization Code grant, so no shared secret is stored anywhere.
* **Per-user audit trail** — Marketing Cloud sees each agent action under the actual user's identity, not a shared API user.
* **Refresh tokens stay server-side** — they're stored in MeshMesh's encrypted vault, never on your device. Short-lived access tokens are refreshed automatically.
**What you need to do**
Existing Marketing Cloud connections will be deactivated and need to be set up again as a new Public App integration. The setup is a one-time admin task in your Marketing Cloud account — creating an Installed Package and adding a Public App API Integration — and then each team member who uses the integration signs in with their own Marketing Cloud account.
→ [Connecting Marketing Cloud Engagement](/guides/connect-marketing-cloud)
## General
* Reliability and security improvements
# April 28, 2026 - Small UX improvements
# April 28, 2026
## What's Improved
* **Celebration no longer blocks the chat** — the task completion card now sits over the left panel instead of the chat, so you can keep scrolling through the conversation while you wrap up.
* **More reliable tag editing** — adding a tag from the suggestions list now registers on the first click, and the sidebar tag filter is a clean single-select: click to filter, click again to clear.
* **More reliable brand kit creation** — brand kits that pull from the web use a steadier browser automation path under the hood.
* Stability and reliability improvements
# May 1, 2026 - Smoother Marketing Cloud reconnection
# May 1, 2026
## What's Improved
* **Marketing Cloud reconnection now Just Works** — if your Marketing Cloud sign-in has expired or been revoked, starting a task now opens the reconnection modal up front, with the existing connection's details already filled in. One click to reconnect, then your task continues — no more cryptic "credentials could not be loaded" mid-task.
* **No duplicate connections on reconnect** — reconnecting an existing Marketing Cloud connection updates it in place instead of creating a second copy, so your connection list stays clean.
* **Long Marketing Cloud agents stay signed in** — long-running tasks that use Marketing Cloud no longer get knocked offline when the agent quietly refreshes its sign-in in the background.
* Stability and reliability improvements
# May 4, 2026 - Tidy up your integrations
# May 4, 2026
## What's New
* **Turn integrations on or off** — every integration in the Your Integrations modal now has a toggle. Switch one off and it disappears from the agent prompt picker, but stays in your account so you can flip it back on whenever you need it.
## What's Improved
* **Delete legacy Marketing Cloud connections in one click** — when you click an old Marketing Cloud Engagement connection, the deprecation prompt now offers Delete and Delete & create new connection directly. No more navigating through menus to clean up old connections.
# May 5, 2026 - MeshMesh Studio on the Web
# May 5, 2026
## What's New
### MeshMesh Studio on the Web
You can now use MeshMesh Studio in your browser at **[studio.meshmesh.io](https://studio.meshmesh.io)** — sign in with the same account you already use. No install required.
This is feature-parity with the desktop app for everyday work: starting agents, picking references, managing connections, browsing the dashboard. The desktop app is still the way to go if you need the local file browser or want everything to keep running while you close your laptop, but for quick sessions, sharing a screen, or a teammate who hasn't installed yet, the browser now works.
The desktop app currently supports macOS only; Windows and Linux remain on the roadmap. The web app works in any modern browser.
### Smoother onboarding setup
The "Connect Your Tools" step in onboarding now uses the same Integrations experience as the rest of the app — so a new teammate can complete the proper Salesforce, Marketing Cloud, Slack, GitHub, or Mulesoft OAuth from inside their first run, with credentials stored server-side just like in the main app.
Connection types are now scrollable when you have many available, deprecated integrations don't appear for new connections (you can still see the ones you already have), and the Marketing Cloud v2 OAuth flow you'd expect from the main app is fully wired through.
### Friendlier credits experience
A few rough edges around credits and warnings:
* **No more false "out of credits" warnings** — the chat now waits until we've actually confirmed your balance before flagging.
* **Low-credits banner** — a gentle heads-up appears before you hit zero, and clears within a second of you topping up.
* **Self-clearing past errors** — if a task hit a credit issue and you've since refilled, the error card adapts back to a normal state instead of staying red.
## What's Improved
* **Marketing Cloud setup guide updated** — the [connection guide](/guides/connect-marketing-cloud) now documents the redirect URI for each surface, and explains how to set things up if your team uses MeshMesh from both desktop and web.
* **Setup guide for the MeshMesh App connection** — selecting "MeshMesh App" when adding a Salesforce connection now links straight to the [setup guide](/guides/connect-salesforce-org#meshmesh-connected-app-pre-release-supported-clients-only) so first-time admins know about the one-time package install.
* **Cleaner integrations popover in chat** — the connection picker in the chat input is now a single, consistent dropdown across the app.
* **Polished connection screens** — the Back button on the connections modal now closes cleanly when you opened it from a contextual link (the chat, the dashboard, onboarding). Dialogs no longer auto-focus their first button on open, so you won't accidentally trigger an action by hitting Enter.
* Reliability and performance improvements.
# May 6, 2026 - Improved connection handling on the web
# May 6, 2026
## What's Improved
### Improved connection handling on the web
Connecting a Salesforce org from MeshMesh Studio in the browser is now more reliable end-to-end. Sign-in opens in a small popup so the connection dialog stays where it is, errors surface inline, and the popup closes itself the moment login finishes.
Cancelled or failed logins no longer leave half-finished connections behind in your list, and retrying after an interrupted attempt picks up cleanly. The desktop app keeps its existing one-click deep-link sign-in — nothing changes there.
### Tutorial videos now play on the web app
The embedded tutorial videos in onboarding, on the dashboard, and in settings now play correctly when you're using MeshMesh Studio in the browser. Previously they were blocked by the browser's content security policy and showed a black box.
# May 7, 2026 - Workbenches & Multi-Agent Collaboration
# May 7, 2026
## What's New
* **Workbenches** — a new shared workspace where multiple agents collaborate on the same project. Set your connections and references once on the workbench, and every agent inside picks them up automatically. Agents can hand work off to each other directly — or @-mention a peer to pull them in mid-task — so you can have one analysing pipeline, another running a marketing report, and a third pulling it all together, all without copying state around.
* **Plan mode by default** for new agents in a workbench, so they map out the work before changing anything. Switch to Build any time you're ready to ship.
* **Watch each agent live** — every agent in a workbench has its own browser view, running in its own isolated session so multiple agents can browse in parallel without stepping on each other. Switching tabs swaps the live view to that agent.
* **Shared workspace + per-agent folders** — each workbench has a shared folder all agents can read and write, plus a private scratch folder per agent so they never overwrite each other's files.
* **Unified Outputs** — see every agent's artifacts from the whole workbench in one place.
* **Polished new-workbench flow** — the create dialog now shows a calm "Crafting" state while your shared workspace is being prepared.
## What's Improved
* Live browser sessions are noticeably faster and smoother — frame-paced rendering, native-feel scrolling on Mac, and the view auto-follows whichever tab the agent jumps to
* New agents in an existing workbench start faster — we skip warm-up steps when the workspace is already running
* New workbench agents now default to our latest model for sharper planning and faster work
* Editable workbench titles, plus right-click actions on agent tabs
* Bigger, more legible dialogs across the app
* The drawer's "Archived" sections only appear when there's actually something to restore, and load on demand so the drawer opens instantly
* Live browser view recovers gracefully if the connection blips, and clears itself when an agent closes the page
* Archiving an agent now asks to confirm, and cleans up coordination state behind the scenes
[Learn about workbenches →](/concepts/workbenches) · [Working with Workbenches guide →](/guides/working-with-workbenches)
# May 8, 2026 - Live Browser Rendering Improvements
# May 8, 2026
## What's Improved
* **Improved page rendering in the agent's live browser** — Salesforce, Trailhead, Marketing Cloud, and other sign-in and signup pages now load cleanly in the live browser view, so you can hand the agent off to a login screen with confidence.
* **Correct aspect ratio in the live browser** — pages now render at the full 1280×800 viewport instead of looking slightly horizontally squished, matching what you'd see in a normal desktop browser.
# May 12, 2026 - Faster, leaner agent browsing
# May 12, 2026
## What's New
* **Faster agent browsing** — when an agent opens the browser to click through Salesforce, navigate docs, or fill in a form, every turn is now noticeably snappier. You'll feel it most on long tasks with lots of clicks, snapshots, and page changes — the back-and-forth between your agent and the page just moves quicker.
* **Up to 75% fewer tokens per browser turn** — the same browser work now uses far less context, so browser-using agents stay leaner over long tasks. Agents that spend a lot of time clicking around in the browser will feel the difference most.
## What's Improved
* Live browser view stays in sync with whichever agent tab you're on, even after long pauses
* More consistent behaviour across single agents and big workbenches with many concurrent browser sessions
* Quieter, more reliable handoff between an agent finishing a task and your next follow-up
* Steadier long-running browser tasks across Salesforce, Marketing Cloud, and the wider web
[Learn about browser use →](/concepts/browser-use) · [Working with Workbenches guide →](/guides/working-with-workbenches)
# May 13, 2026 - More reliable Salesforce connections
# May 13, 2026
## What's Improved
* **Salesforce connections stay signed in** — fixes a bug where multiple Salesforce connections in the same org (especially ones sharing the same Connected App, such as the MeshMesh App Package or a shared ECA) could repeatedly fail with "Connection credentials could not be loaded. Please reconnect." even after recreating the connection. Existing affected connections will need one final reconnect, after which they should stay healthy.
* **Custom Connected App** — now correctly only available in the desktop app, and fixes an issue where some Custom Connected App connections weren't resolving properly on the desktop.
[Salesforce connection guide →](/guides/connect-salesforce)
# May 15, 2026 - Cloud Control 5.1 launches in preview
# May 15, 2026
## What's New
* **Cloud Control 5.1 (Preview) — our largest context model yet.** Now available in preview. CC5.1 expands and deepens coverage across:
* **Health Cloud, Education Cloud, Public Sector Solutions, Nonprofit Cloud, Consumer Goods Cloud, OmniStudio, Revenue Cloud / CPQ, CRM Analytics, Tableau Next, Marketing Intelligence** — now covered.
* **Agentforce** — Agent Script (`.agent` files + `sf agent` CLI) is now the recommended authoring path for custom-action agents. Deeper coverage of deployment and web deployment, event-log debugging, bot configuration, Builder UI, GenAI Functions / Plugins / Planners, prompt templates, and Apex services.
* **Data Cloud** — web SDK, TypeScript SDK, data graphs, data spaces, transforms, query API, real-time actions, clean rooms, scoring models, ingestion depth, knowledge and search.
* **Platform** — Apex patterns, browser automation, bulk API, CI/CD, custom fields, metadata deployment, page layouts, permissions, Platform Events, SOQL, approval processes, validation rules, reports.
* **Field Service** — Dispatcher Console, Service Resources, Asset Attributes, Parts Consumption.
* **Commerce Cloud** — inventory across LocationGroups, full OMS with Connect API, order-on-behalf, product visibility and entitlements.
* **Experience Cloud** — CMS workspace permissions, Embedded Messaging security setup, deployment patterns.
* **Marketing Cloud Engagement** — Campaign Flows, Distributed Marketing, MobileConnect two-way SMS.
* **Service Cloud** — Command Center, Next Best Action, social media response patterns.
* **Lightning Apps, Flows, Loyalty Management** — Path Assistant, theming, browser automation, quick promotions.
* **CC5.1 is the new default for new agents** — dashboard quick-create and the workbench "+" both preselect CC5.1 (Preview). Switch between CC4 and CC5.1 in the Cloud Control popover when starting an agent.
## What's Improved
* **MeshMesh App now works for sandboxes and custom My Domain orgs** — connecting a Salesforce sandbox or custom My Domain org via the **MeshMesh App** method no longer fails with "Grant type not supported". The OAuth token exchange now routes to the matching Salesforce server based on the **Environment** you select in the Connections modal.
* **Install link matches your environment** — the **Install MeshMesh Studio** button in the "MeshMesh Studio Not Installed" prompt now opens the install URL for the org you're trying to connect to (production, sandbox, or your custom My Domain), instead of always sending you to production.
[Salesforce connection guide →](/guides/connect-salesforce-org)
# May 19, 2026 - Marketing Cloud reconnect loop fixed, browser typing fixes
# May 19, 2026
## What's New
* **Fixed Marketing Cloud "please reconnect" loop on long tasks** — agents 20+ min into a Marketing Cloud run no longer spuriously ask you to refresh the connection
* **Correct connection picker when a Marketing Cloud refresh is genuinely needed** — shows the right connection type and only matching connections, not every connection in your workspace
* **Auto-recovery from one-off Marketing Cloud sign-in errors** — agents self-heal from transient token hiccups without interrupting the task
* **Fixed typing special characters into the agent's browser** — `@`, `!`, `#`, `$` and other shifted symbols sometimes got dropped on login forms
* **Fixed browser thumbnail overlay overlapping workbench tabs** — the "Agent browsing" indicator now lives in the page header with a working live preview
## What's Improved
* Multi-agent setups on the same Marketing Cloud connection coordinate cleanly
* Better internal logs for Marketing Cloud connection issues
* Browser thumbnail keeps updating while you're on the Resources tab
[Connect your Marketing Cloud →](/guides/connect-marketing-cloud) · [Working with Workbenches guide →](/guides/working-with-workbenches)
# May 26, 2026 - Usage limits with a clear heads-up, plus a new Admin Console
# May 26, 2026
## What's New
* **Monthly usage limits with a clear heads-up** — workspaces can now set a monthly usage limit. As you approach it you get a gentle in-app warning, and if you reach it you see a clear message instead of a task quietly failing
* **See your usage in the app** — a new usage view shows your workspace's monthly usage at a glance, so there are no surprises
* **Admin Console for workspace owners** — members, usage limits, allowed integrations, and an audit log of workspace activity are now managed in one place. Deactivated members now show up in the members list too
## What's Improved
* Usage is now attributed per person, so admins can see who's using what across the workspace
# May 27, 2026 - Steadier desktop sign-in, a minimizable review card, better Agentforce builder detection
# May 27, 2026
## What's New
* **Review your agent's work without losing your place** — the task-complete review card can now be minimized into a small "Review Agent" pill in the corner, so it no longer covers the references and artifacts you're trying to look over. Click the pill to bring it back, or choose "I'm not done yet" to dismiss it for now
* **Steadier sign-in on desktop** — hardened a few edge cases (like overlapping background refreshes) where the desktop app could occasionally drop your session; it now refreshes quietly and retries before signing out
## What's Improved
* **Better Agentforce builder detection** — fixed some cases where agents would reach for the legacy Agentforce builder; they now more reliably use the new Agentforce Builder (Agent Script) when that's what you want, and ask which you'd like when it's ambiguous
* The review card resets cleanly when you re-run an agent, so the next turn's review opens fresh
# May 29, 2026 - Smoother member management, more reliable Salesforce & Marketing Cloud connections
# May 29, 2026
## What's New
* **Smoother member management in the Admin Console** — inviting members now works reliably, with clear, actionable messages when someone has already been invited or is already a member. You can also search your member list by name or email, re-invite people with a pending invitation, and act on several members at once with new bulk actions (invite, deactivate, reactivate, and revoke)
* **A tidier Admin Console** — options that aren't enabled for your workspace no longer appear where they can't be used
## What's Improved
* **Fixed Salesforce agents getting stuck asking to reconnect** — resolved an issue affecting both **Quick Connect** and **Custom App** Salesforce connections, where agents could get caught in a loop, repeatedly asking you to reconnect and reporting that no orgs were found, even when the connection was healthy. Agents now pick up your connection reliably, and you'll see a clear message if one ever genuinely needs attention
* **Smoother Marketing Cloud reconnection on the web** — when you reconnect Marketing Cloud in your browser, the sign-in window now opens and closes itself cleanly, and no longer reports success before you've actually finished signing in
* **Security & reliability improvements** — behind-the-scenes hardening across the platform, including tighter protection of our services, dependency security patches, and a reduced log surface, alongside a large expansion of automated test coverage. No change to how you work day to day
# May 30, 2026 - Task ID in Salesforce webhooks, behind-the-scenes reliability
# May 30, 2026
## What's New
* **Task ID in Salesforce webhooks** — Salesforce webhook payloads now include the MeshMesh task ID, so a triggered task is easy to tie back to its source
## What's Improved
* Quieter handling of transient sandbox hiccups
* Improved Data Cloud guidance for agents after a recent Salesforce CLI change
* A large expansion of automated test coverage
[Connect Salesforce →](/guides/connect-salesforce-org)
# June 1, 2026 - Richer Activity Reports and smarter, more up-to-date agents
# June 1, 2026
## What's New
* **Richer Activity Reports** — task cards now show when they were last updated and their star rating, the leaderboard adds an Agent Time column, and Active Teammates counts everyone who has run an agent
* **Smarter, more up-to-date agents** — agents now check the latest product docs before reaching for a skill, so they apply the freshest guidance and give more accurate answers
## What's Improved
* More consistent task cards across the activity feed and user profiles
* Clearer Activity hero stats and leaderboard ranking
# June 2, 2026 - Business Intelligence loads cleanly, correct invite attribution, hardened sign-in
# June 2, 2026
## What's New
* **Business Intelligence opens cleanly** — opening Business Intelligence now shows a brief loading state and then your existing analyses, instead of flashing "No analyses yet" before they load
* **Correct inviter on team invitations** — invitations now record and show the right person as the inviter
## What's Improved
* Hardened desktop sign-in behind the scenes — nothing changes about how you sign in
* Dependency security patches across the stack
# June 7, 2026 - Reliable Salesforce connections, accurate Activity numbers
# June 7, 2026
## What's New
* **More reliable Salesforce connections — with a clear way back** — a failed Salesforce connection used to leave an agent stuck "searching for the connection." Now a failed task shows a banner with a one-click **Reconnect** button, and the agent gets back to work as soon as you reconnect
* **Reconnect badges where you need them** — connections that need a refresh show an amber badge in your Integrations list and in the task connection picker, so you can spot and fix them at a glance
* **Accurate Activity user counts** — the Activity dashboard's total-users count now reflects every active member of your workspace, instead of capping out and under-counting on larger teams
* **Leaderboard updates reliably** — the Activity leaderboard refreshes its stats dependably again
## What's Improved
* Behind-the-scenes security hardening
* Dependency security patches across the stack
[Connect Salesforce →](/guides/connect-salesforce-org)
# June 12, 2026 - Live-updating file browser and steadier performance under load
# June 12, 2026
## What's New
* **Live-updating file browser** — the agent's file browser now refreshes the moment files change, using a real-time signal instead of polling every few seconds, so new and edited files appear without the wait
* **Marketing Cloud connections stop spuriously asking to reconnect** — fixed a reauthentication loop that could repeatedly prompt some Marketing Cloud connections to reconnect; they now refresh quietly, matching the Salesforce fix from June 7
## What's Improved
* Fewer timeouts and failed task setups during busy periods — behind-the-scenes database improvements keep tasks starting, resuming, and browsing files reliably under heavy load
* Behind-the-scenes security hardening
* Dependency security patches across the stack
[Connect Marketing Cloud →](/guides/connect-marketing-cloud)
# June 16, 2026 - Faster startup and load times, manage task outputs, steadier confirmations
# June 16, 2026
## What's New
* **Faster startup and load times** — improved startup performance on the desktop app, so it gets you to your tasks and workspace quicker on refresh or relaunch, with the latest data filling in as it loads
* **Remove a task's outputs** — you can now delete outputs (artifacts) from a task, just like references, to keep a task tidy
## What's Improved
* Fixed a rare screen freeze when confirming a sensitive action — the confirmation now always stays reachable, with a clear way to dismiss it
* High-severity dependency security patches and behind-the-scenes hardening
# July 22, 2026 - A better-organized library for your references and outputs
# July 22, 2026
This release makes the references and outputs in an agent's workbench far easier to organize and act on — clearer lists, consistent actions everywhere, and folders to group things your way.
## What's New
* **A clearer Generated Outputs list** — outputs now show as a sortable, newest-first list with the time each was created and which agent produced it, so two similarly-named outputs are easy to tell apart. Sort by newest, name, or type and filter by type — your choice sticks per workbench
* **Consistent actions on every item** — hover or right-click any reference or output, in list or grid, to rename, star, download, remove, or save it to your library. Actions work the same way everywhere
* **Save an output to your library** — turn a great output into a reusable reference in a couple of clicks, instead of asking the agent to recreate it in chat
* **Folders for references and outputs** — group a task's items into folders, Finder-style: create a folder, drag items in, or right-click to add and remove; open a folder to focus on its contents and use the breadcrumb to step back. Empty folders are fine — make one first and fill it later
* **Download on the web, and in bulk** — the output viewer now downloads on the web too, not just desktop. Select several items and download them together as a single ZIP, and zipped exports (like metadata bundles) download intact
* **Find within a document (⌘F)** — search inside a plan, reference, or output to jump between highlighted matches with a running count — handy for QA-ing a long plan. Esc closes, Enter cycles
## What's Improved
* **A cleaner references picker** — the add-references picker now shows each reference's description and how many tasks use it, so duplicates are easy to distinguish. Searching with a type filter active no longer hides real matches, and your filters persist between opens
* **Capture details when you save a reference** — the new-reference dialog now lets you add a description, type, and tags up front, so your library stays organized from the start
* **Fewer near-duplicate outputs** — when an agent refines work it already produced, it now updates that output instead of filing a new "v2" each time, keeping your Outputs manageable
* **Plans appear in one place** — a plan now shows once, under Generated Outputs, instead of also appearing under Resources
* High-severity dependency security patches and behind-the-scenes hardening
# July 24, 2026 - Connect Salesforce with your own External Client App
# July 24, 2026
This release adds a new way to connect Salesforce Core using your own External Client App, so your agents can run on the web and keep working on their own — no desktop app needed. It's especially handy if you're on Windows and rely on the web version.
## What's New
* **Connect Salesforce with your own External Client App** — a new connection method under Salesforce Core lets you connect using an External Client App you create in Salesforce. Because the connection runs on our servers, your agents work on the web and can keep running on their own — including scheduled and always-on tasks — without the desktop app open. This is especially useful if you're on Windows and work in the web version, where always-on agents now work the same as on desktop. Paste in your app's Consumer Key, copy the callback URL we give you straight into Salesforce, and you're connected. Your app's own settings decide exactly what the agent can access.
## What's Improved
* Security patches and fixes, plus behind-the-scenes hardening
# July 28, 2026 - More reliable output downloads, plus stability and security fixes
# July 28, 2026
This release fixes an issue that could cause output downloads to fail, and includes behind-the-scenes stability and security fixes.
## What's Fixed
* **Fixed an issue where output downloads could fail** — downloads previously depended on the task's live workspace, so they could fail with a "Failed to download file" message when the task had gone idle or asleep. Outputs now render from your saved content instead, making downloads far more reliable
* **Fewer dropped messages during very busy tasks** — fixed a database issue that could, under heavy load, occasionally fail to record a message and quietly lose it
## What's Improved
* **Security & reliability improvements** — dependency security patches and behind-the-scenes hardening
# August 4, 2026 - A sign-in loop fix for newly invited members, plus reliability and security fixes
# August 4, 2026
This release fixes a sign-in loop that could affect some newly invited members, helps the desktop app stay on the workspace you chose, and includes security and reliability improvements.
## What's Fixed
* **Fixed a sign-in loop that could affect newly invited members** — in some cases, an invited teammate who had been added but hadn't finished accepting could be signed out again shortly after signing in, until an admin toggled their access. Invited members now stay signed in and reach their workspace the first time
* **Better at keeping you in the workspace you selected** — in some cases, switching workspaces on desktop could snap you back to the one you signed in with; your chosen workspace now stays put more reliably, even when the app refreshes your sign-in in the background
## What's Improved
* **Security & reliability improvements** — dependency security patches and behind-the-scenes hardening
# August 24, 2026 - Connect MCP servers to your agents
# August 24, 2026
This release lets you connect remote MCP servers to your agents with a secure sign-in, makes Salesforce connections more reliable, and includes security and reliability improvements.
## What's New
* **Connect your favorite tools to agents over MCP** — connect remote MCP servers with a secure sign-in (OAuth) and let your agents use them directly in a task. Popular servers work out of the box — **Linear, Notion, Atlassian, Tableau, and Heroku** — and you can connect **any custom MCP server** by pasting its URL. Connect once, give each connection a clear name, and turn connections on or off whenever you like
## What's Fixed
* **More reliable Salesforce connections** — if a Salesforce connection can no longer refresh its access, MeshMesh now clearly asks you to reconnect and explains why, instead of quietly failing partway through a task
## What's Improved
* **Security & reliability improvements** — some behind-the-scenes security hardening
# August 25, 2026 - Faster, more reliable Analytics
# August 25, 2026
This release makes the Analytics section faster and more reliable — especially on larger workspaces — and adds a few improvements to how you view and pull member information.
## What's Improved
* **Faster, more reliable Analytics** — the Analytics → Users view now loads from precomputed usage data instead of calculating everything on the fly, so it opens quickly and reliably even on large workspaces. Previously it could time out and show "No users yet" even when you had plenty of activity
* **Member roles and join dates** — the Users view now shows each member's role and when they joined
* **Pull your full member roster with agents** — a new admin capability lets agents retrieve your complete workspace member list, including people who haven't been active yet
## What's Fixed
* **Clearer errors instead of an empty state** — if the Analytics view can't load, you now see a clear message with a **Retry** button, instead of it silently showing "No users yet"
# August 26, 2026 - Reach systems behind enterprise IP allowlists
# August 26, 2026
This release lets your agents reach systems that sit behind enterprise IP allowlists.
## What's New
* **Reach systems behind enterprise IP allowlists** — agents now connect through stable, dedicated outbound IP addresses. That means they can work with systems that only accept traffic from approved addresses, including **GitHub Enterprise**, **Marketing Cloud**, and **IP-restricted Salesforce orgs**. Your IT team allowlists MeshMesh's outbound addresses once, and agents can securely reach those systems from then on
# Analysis Patterns
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
These patterns show how to ask for analysis, audits, and insights. Start with these examples and adapt them to your needs.
## Email Performance Analysis
**Basic analysis:**
```
Analyze my email performance for the last 30 days and identify
which campaigns had the lowest engagement
```
**With comparison:**
```
Compare email performance across all campaigns in Q4. Show me
what's working and what's not, with specific recommendations.
```
**Trend analysis:**
```
Show me how our email open rates have changed month over month
this year. Are we trending up or down? What's driving the trend?
```
**Segment-level:**
```
Break down email engagement by customer segment. Which segments
respond best to which types of content?
```
## Pipeline Analysis
**At-risk deals:**
```
Analyze my sales pipeline and identify deals at risk of slipping.
What are the warning signs and what should we do about them?
```
**Pipeline health:**
```
Create a pipeline health report showing conversion rates at each
stage. Where are we losing deals?
```
**Forecasting:**
```
Based on historical patterns, predict which open opportunities
are most likely to close this quarter.
```
## Tech Debt Audits
**Marketing Cloud:**
```
Run a tech debt audit on my Marketing Cloud account. Find unused
data extensions, broken journeys, and outdated email templates.
Prioritize what to clean up first.
```
**Salesforce org:**
```
Audit my Salesforce org for technical debt. Look for unused custom
fields, outdated automation, and duplicate processes.
```
**Data Cloud:**
```
Audit my Data Cloud setup. Are there data streams that haven't
run recently? Segments that aren't being used? Identity resolution
rules that might be outdated?
```
Tech debt audits work best when you ask for prioritization. "What should I clean up first?" helps focus on high-impact changes.
## Data Quality Assessment
**Contact data:**
```
Analyze the quality of our contact data. What percentage have
valid emails? How many duplicates do we have? What's missing?
```
**Data completeness:**
```
Check how complete our customer records are. For each key field,
show me what percentage of records have data.
```
**Consistency:**
```
Look for data consistency issues across Sales Cloud and Marketing
Cloud. Are customer records in sync? Where are the gaps?
```
## Security and Compliance
**Access review:**
```
Audit user permissions in my Salesforce org. Who has admin access?
Are there users with permissions they don't need?
```
**Sharing rules:**
```
Review our sharing settings. Is sensitive data properly restricted?
Are there any overly permissive rules?
```
**Field-level security:**
```
Check field-level security for our custom objects. Are PII fields
properly protected?
```
## Performance Reporting
**Executive summary:**
```
Create an executive summary of our marketing performance this
quarter. Include key metrics, wins, and areas for improvement.
Keep it to one page.
```
**Dashboard creation:**
```
Build a dashboard showing our key sales metrics: pipeline value,
win rate, average deal size, and sales cycle length.
```
**Attribution analysis:**
```
Analyze which marketing channels are driving the most revenue.
Include first-touch and last-touch attribution.
```
## Customer Analysis
**Churn prediction:**
```
Identify customers at risk of churning. Look at engagement drop-off,
support tickets, and usage patterns.
```
**Lifetime value:**
```
Calculate customer lifetime value by segment. Who are our most
valuable customers and what do they have in common?
```
**Behavior patterns:**
```
Analyze customer behavior patterns before their first purchase.
What actions predict conversion?
```
## Configuration Analysis
**Flow review:**
```
Review all active Flows in my org. Which ones have errors?
Which ones haven't fired in 90 days?
```
**Automation overlap:**
```
Check for automation that might be overlapping or conflicting.
Are there multiple automations triggered by the same events?
```
**Integration health:**
```
Review our integrations with external systems. Are they running
successfully? Any errors or failures?
```
## How to Get Better Analysis
**Be specific about time ranges:**
```
Analyze email performance for Q4 2024, not just "recently"
```
**Ask for comparisons:**
```
Compare this quarter to last quarter, not just show current metrics
```
**Request recommendations:**
```
Don't just show me the data — tell me what to do about it
```
**Prioritize findings:**
```
If there are many issues, tell me which ones to fix first
```
## Next Steps
See example prompts for building and deploying
See example prompts for campaigns and operations
# Implementation Patterns
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
These patterns show how to ask for implementations across Salesforce products. Start with these examples and adapt them to your needs.
For complex implementations, start in **Plan mode** to get an implementation plan before building. See [Plan Then Build](/docs/prompting/plan-then-build).
## Sales Cloud
**Lead scoring:**
```
Implement lead scoring in Sales Cloud based on engagement signals
from Marketing Cloud. Score leads based on email opens, clicks,
and website visits.
```
**Case routing:**
```
Create a Flow that routes cases to the right queue based on
product category and customer tier.
```
**Opportunity automation:**
```
Build automation that updates opportunity stage when key tasks
are completed. Send alerts to managers when deals are stuck.
```
**Custom objects:**
```
Create a custom object to track partner referrals. Include fields
for partner name, referral date, opportunity link, and commission
status.
```
## Service Cloud
**Case management:**
```
Implement a case escalation process. Cases should auto-escalate
after 24 hours without response for standard customers, 4 hours
for premium.
```
**Knowledge base:**
```
Set up a knowledge base structure for our support team.
Categories should match our product lines. Include article
templates for FAQs, troubleshooting, and how-to guides.
```
**Service metrics:**
```
Create reports and dashboards for service metrics: average
resolution time, first contact resolution rate, CSAT by agent.
```
## Data Cloud
**Unified profiles:**
```
Create a unified customer profile in Data Cloud using data from
Sales Cloud and Marketing Cloud. Use email as the primary identity
key with fallback to phone number.
```
**Segments:**
```
Create a segment for high-value customers who haven't engaged
in the last 30 days. High-value means lifetime spend over $5000.
```
**Data streams:**
```
Set up a data stream to ingest product data from our external
inventory system via IngestAPI.
```
**Activation:**
```
Configure an activation target to export our VIP segment to an
S3 bucket daily for use by our data warehouse.
```
For Data Cloud implementations, Plan mode is especially valuable. The agent can explore your existing data model and connections before proposing changes.
## Marketing Cloud
**Email journey:**
```
Create a welcome journey for new subscribers. Send a welcome email
immediately, a product showcase after 3 days, and a discount offer
after 7 days if they haven't purchased.
```
**Data extensions:**
```
Create a data extension to store lead scores synced from Sales
Cloud. Fields: email, score, last_calculated, engagement_tier.
```
**Content blocks:**
```
Create reusable content blocks for our standard email header,
footer, and product grid. Apply our brand kit.
```
**Automation:**
```
Set up an automation that runs daily to identify subscribers who
haven't opened an email in 90 days and add them to a re-engagement
segment.
```
## Agentforce
**Service agent:**
```
Build an Agentforce service assistant that can look up orders,
check shipping status, and create return requests. Connect it
to our Service Cloud org.
```
**Sales assistant:**
```
Create an Agentforce sales assistant that can retrieve customer
history, suggest next best actions, and create follow-up tasks.
```
**Custom functions:**
```
Add a new function to my existing agent that calculates customer
loyalty points based on their purchase history.
```
## Flows
**Screen flow:**
```
Create a screen flow for new customer onboarding. Collect company
info, primary contact, and preferences. Create Account, Contact,
and a custom onboarding record.
```
**Record-triggered:**
```
Build a record-triggered flow that fires when an opportunity
closes won. Create a customer success case and notify the
assigned CSM.
```
**Scheduled:**
```
Create a scheduled flow that runs weekly to identify accounts
without activity in 60 days and creates renewal reminder tasks.
```
## Apex
**Integration:**
```
Write an Apex class that syncs contact data to our external
marketing platform via REST API. Include error handling and
retry logic.
```
**Trigger:**
```
Create a trigger on the Lead object that validates phone number
format and auto-formats to E.164 standard.
```
**Batch:**
```
Build a batch Apex class to recalculate customer health scores
nightly based on engagement metrics.
```
## Permission Sets
**Feature access:**
```
Create a permission set for our sales team that grants access
to the new opportunity fields and the customer health dashboard.
```
**Object access:**
```
Create a permission set for partners that gives read-only access
to their referred opportunities and commissions.
```
## How to Get Better Implementations
**Start with the outcome:**
```
I want customers to automatically move to a VIP segment when
their lifetime value exceeds $10,000
```
**Include constraints:**
```
This needs to work with our existing lead routing. Don't duplicate
or override the current process.
```
**Specify the org:**
```
Deploy this to my dev sandbox first. I'll review before we move
to production.
```
**Request testing:**
```
After building, test with sample data to verify it works correctly.
```
## Next Steps
See example prompts for campaigns and operations
Learn the recommended implementation workflow
# Operate Patterns
import { Callout } from 'fumadocs-ui/components/callout';
import { Card, Cards } from 'fumadocs-ui/components/card';
These patterns show how to ask for operational tasks — running campaigns, maintaining data, and managing ongoing processes.
## Campaign Execution
**Re-engagement:**
```
Create a re-engagement campaign for customers who haven't purchased
in 90 days. Offer a 20% discount and personalize based on their
last purchase category.
```
**Win-back:**
```
Build a win-back campaign for churned customers. Segment by reason
for leaving (if known) and customize messaging accordingly.
```
**Upsell:**
```
Create an upsell campaign targeting customers who've purchased
our starter product. Highlight the benefits of upgrading to pro.
```
**Seasonal:**
```
Set up our Black Friday campaign. Target active customers with
a VIP early access offer, then open to general list on Friday.
```
## A/B Testing
**Subject lines:**
```
A/B test subject lines for our newsletter. Test a question-based
subject against a benefit-focused one. Run for one week and
implement the winner.
```
**Send times:**
```
Test sending our weekly digest on Tuesday vs Thursday. Track
open rates and clicks for 4 weeks.
```
**Content:**
```
A/B test our welcome email. Version A has a discount, Version B
has educational content. Which drives more first purchases?
```
## Data Maintenance
**Duplicate cleanup:**
```
Find and merge duplicate contacts in my org. Match on email
address first, then name + company as a secondary match.
```
**Data enrichment:**
```
Identify contacts missing job titles. Flag them for manual
enrichment or mark as low priority.
```
**List hygiene:**
```
Clean up our email list. Remove hard bounces, identify inactive
subscribers (no engagement in 180 days), and segment for
re-engagement or removal.
```
**Field updates:**
```
Bulk update all contacts with @oldcompany.com email addresses
to @newcompany.com — they rebranded.
```
For bulk data operations, always ask the agent to show you a preview before making changes.
## Reporting and Monitoring
**Weekly reports:**
```
Set up a weekly email performance report that sends to the
marketing team every Monday. Include opens, clicks, conversions,
and comparison to previous week.
```
**Pipeline alerts:**
```
Create an alert that notifies sales managers when a deal over
$50k hasn't had activity in 14 days.
```
**Health monitoring:**
```
Set up monitoring for our Marketing Cloud automations. Alert me
when any automation fails or hasn't run on schedule.
```
**Executive dashboard:**
```
Build a monthly executive dashboard showing revenue attribution,
customer acquisition cost, and lifetime value trends.
```
## Segmentation
**Behavioral:**
```
Create a segment of customers who've viewed our pricing page
3+ times but haven't started a trial.
```
**RFM:**
```
Build segments based on recency, frequency, and monetary value.
Create 5 tiers from VIP to at-risk.
```
**Engagement:**
```
Segment our email list by engagement level: highly engaged
(opens everything), moderate (opens sometimes), and cold
(hasn't opened in 60+ days).
```
**Lifecycle:**
```
Create lifecycle segments: new customers (0-30 days), growing
(30-90 days), established (90+ days with multiple purchases),
at-risk (no activity in 60 days).
```
## Content Operations
**Template updates:**
```
Update all our email templates to use the new footer with
updated legal copy.
```
**Asset organization:**
```
Organize our Content Builder assets into folders by campaign
type: newsletters, promotions, transactional, and lifecycle.
```
**Content calendar:**
```
Create a content calendar for Q1. Plan one newsletter per week,
two promotional campaigns, and the quarterly product update.
```
## Integration Management
**Sync verification:**
```
Verify that our Sales Cloud to Marketing Cloud sync is working
correctly. Check for records that didn't sync and identify why.
```
**Error resolution:**
```
Review errors from yesterday's data import. Tell me what failed
and why, then fix what can be fixed automatically.
```
**Connection testing:**
```
Test all our Data Cloud connections. Are they still authenticated?
When did they last sync successfully?
```
## Automation Management
**Pause and resume:**
```
Pause all promotional automations for the holiday period
(Dec 24-26). Resume on Dec 27.
```
**Schedule adjustments:**
```
Move our weekly newsletter send from Tuesday at 9am to
Wednesday at 10am, starting next week.
```
**Cleanup:**
```
Identify automations that haven't run successfully in 30 days.
Create a list for review with last run date and error status.
```
## How to Ask for Operations
**Be specific about timing:**
```
Send the campaign at 10am EST next Tuesday, not just "next week"
```
**Include safety checks:**
```
Show me a preview of who will receive this before sending
```
**Request confirmation:**
```
After running the data cleanup, give me a summary of what changed
```
**Set up monitoring:**
```
Alert me if this campaign's unsubscribe rate exceeds 1%
```
## Next Steps
Review the core philosophy of effective prompting
See example prompts for analysis and audits