Connect Claude and Other AI Assistants to Tesselith (MCP)

Tesselith runs a remote Model Context Protocol server, so AI assistants can work in your workspace directly. Ask Claude to "queue three posts about our launch for next week" and they appear in your queue as real posts, each with a link to open it in Tesselith. The server URL is https://api.tesselith.com/mcp. You sign in with your Tesselith account and pick the workspace and permissions, and the assistant can never do more than your own role allows. MCP is included on the Pro, Team and Agency plans.

Set up in Claude (web and desktop)

1. In Claude, open Settings → Connectors and choose Add custom connector.
2. Name it Tesselith and paste https://api.tesselith.com/mcp as the URL.
3. Click Connect. A Tesselith window opens. Sign in if you aren't already.
4. On the consent screen, pick the workspace, untick anything you don't want the assistant to do, and click Allow access.
5. Back in Claude, enable the connector in a conversation and ask it what's in your queue.

Set up in Claude Code

Add the server, then run /mcp inside Claude Code to sign in through your browser:

claude mcp add --transport http tesselith https://api.tesselith.com/mcp

On a server or in CI, where nobody can click through a browser sign-in, create an API key in Settings → Developers with only the scopes the job needs and pass it as a header:

claude mcp add --transport http tesselith https://api.tesselith.com/mcp \
  --header "Authorization: Bearer tsl_live_…"

Other assistants and editors

Any client that supports remote MCP servers over HTTP works: ChatGPT connectors, Cursor, VS Code, Windsurf, and agent frameworks built on the official MCP SDKs. Give it the server URL. It discovers the sign-in automatically: the server answers unauthenticated requests with a pointer to its OAuth metadata, and modern clients identify themselves with a Client ID Metadata Document. Older clients that register dynamically are supported too.

{
  "mcpServers": {
    "tesselith": {
      "type": "http",
      "url": "https://api.tesselith.com/mcp"
    }
  }
}

What the assistant can do

Every tool returns structured JSON with an app_url, so the assistant can link you straight to the post or queue in Tesselith. Read-only tools are marked read-only, and publish_now and delete_post are marked destructive, so clients that support it can auto-approve reads and always ask before anything public happens.

Permissions and safety

• Never more than you. The assistant acts as you, in the workspace you picked, with the permissions you approved, limited by your current role. If an owner changes your role, the assistant's access changes with it on its next request.
• Ask again for more. If you didn't grant a permission and the assistant tries a tool that needs it, the server asks your client to request it. You'll see the consent screen again rather than a silent failure.
• Your social accounts stay private. The assistant never sees network passwords or tokens; Tesselith publishes on its behalf.
• Everything is logged. Actions taken through MCP appear in the workspace audit log, marked as MCP, alongside app and API activity.
• One click to disconnect. Settings → AI assistants lists every assistant you've connected to the workspace with its permissions. Disconnect revokes its tokens immediately.

Example prompts

• "What's going out on Instagram this week? Any gaps?"
• "Draft three posts announcing our webinar on Thursday for LinkedIn and Bluesky, then add them to the queue."
• "Turn this blog post into a five-post Threads series and schedule one a day starting Monday at 9am New York time."
• "Which channel needs reconnecting?"
• "Move my Friday post to next Tuesday at 2pm."

Limits

Each connection can make 120 requests per minute on Pro, 300 on Team and 600 on Agency; API keys share their limit with the REST API. Posts scheduled through MCP count toward your plan's monthly post allowance like any other. The assistant works in one workspace per connection. Connect again and pick another workspace to use a second one.