Remote integration · MCP

TODO for AI from your AI client

Start TODOs, send follow-ups and read results from an MCP client. Agents run in your TODO for AI workspace, with their saved permissions and normal account billing.

Add a remote MCP server with this URL, then sign in and approve access:

https://api.todofor.ai/mcp

Transport: Streamable HTTP. No local server or package to install.

Connect a client

In Claude or ChatGPT, add a custom remote connector if your account supports it. In Cursor or VS Code, add a remote HTTP MCP server. Use the URL above and complete the browser sign-in.

Claude Code

claude mcp add --transport http todoforai https://api.todofor.ai/mcp

Clients with stdio configuration only

Use mcp-remote as a bridge. Requires Node.js. It opens the browser for sign-in.

{
  "mcpServers": {
    "TODOforAI": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://api.todofor.ai/mcp"]
    }
  }
}

For headless clients, use a personal API key from Settings as an Authorization: Bearer <API_KEY> header. Keep it out of shared configuration and source control. API keys grant account access, not a narrower MCP-only scope.

Six tools

run_todo: Start a TODO with a saved agent. Returns immediately.
send_message: Send a follow-up to an existing TODO.
get_todo: Read status, recent messages and the latest reply.
list_todos: List TODOs in a project.
list_projects: Find your projects.
list_agents: Find your saved agents.

Runs are asynchronous. Use get_todo to check progress. Open the returned TODO link to approve actions or complete sign-ins. MCP does not bypass those prompts.