Run your studio from Claude

Connect it in three steps

1. Add the connector

   In Claude (claude.ai or Claude Desktop), add a custom connector with your studio's address followed by /mcp. Any MCP client that speaks streamable HTTP works the same way.

2. Sign in on your studio's own page

   The client finds the sign-in by itself and sends you to your studio's sign-in page. Your password stays with your studio; the client gets a session, checked the same way as in the browser.

3. Ask

   “Which tickets are open?”, “Give the flour warning to an agent.”, “What decisions are waiting for me?” Claude calls the studio's tools and tells you what came back.

Everything you can do, Claude can too

Each action in the studio is one tool, declared once: the window uses it, and so does Claude. Nothing is only in one of the two.

• Files and codeRead, write and search the files in your studio, and build and run a program with the compiler's exact message back.
• AgentsStart an agent on a task, ask it something, read what it did, see what it used, stop it.
• TicketsCreate, find, claim and update tickets, log work and add comments.
• Your organisationIts status and roles, the decisions that wait for you, deciding them, and the emergency stop.
• Language helpSearch the language documentation, its examples and its common mistakes, one answer at a time.

Build MCP servers of your own

1. An MCP server in a few lines

   The language has MCP built into its library: declare your tools in one block, and a program answers MCP over HTTP. The studio comes with a worked example.

2. Published, with sign-in

   Publish the program and it gets its own address. Choose Only me, add that address in Claude, and sign in with your studio login: your own MCP server, running on its own.