Skip to main content

MCP Server

Caesell exposes an MCP (Model Context Protocol) server, so external AI tools and coding agents — Claude Code, Claude Desktop, IDE assistants, or any other MCP client — can read and act on your project's data directly, using the standard MCP protocol instead of calling Caesell's REST API by hand.

Authentication​

Connecting an MCP client doesn't use an API key. Instead, the server gates access behind a normal browser login to your Caesell account — the same login the webapp itself uses. When your MCP client connects, it opens a login page; once you sign in, your own access token is used for every call the client makes, so Caesell always sees the real, authenticated user behind the request rather than a shared service account.

What you can do through it​

Once connected, an MCP client can:

  • List and read projects, boards, and tasks
  • Create, update, and delete tasks and task comments
  • List backlog, sprint, and release tasks
  • Read and manage contexts and team members
  • Assign the Coder or Tester agent to a task
  • Fetch attachment download links
  • Query your project's knowledge base

Because every call carries your own token, Caesell's normal role-based permissions still apply — an MCP client can't do anything your account isn't already allowed to do.

Connecting​

Your Caesell MCP server URL is available from your project admin (it typically ends in /mcp). Add it to your MCP client's configuration the same way you would any other remote MCP server, then complete the browser login when prompted. From then on, the agent you're working with in that tool can pull live context from Caesell — current tasks, board state, team members — directly into its own workflow.