Cursor task manager: give Cursor's agent a task queue over MCP
How to connect Cursor to Done over MCP with mcp.json, what Cursor's agent can do with your tasks, and why it still can't close a gated task on its own.

By Alejandro Rioja5 min read
To give Cursor's agent a task queue, add your task manager to Cursor as a remote MCP server in mcp.json. With Done that is one entry pointing at https://trydone.app/mcp in ~/.cursor/mcp.json, then a sign-in under Customize in Cursor's sidebar. After that the agent can pull the next task from your queue, work it, comment on it and propose completion, and anything you haven't marked auto-OK still waits for your approval before it counts as done.
How do I add Done to Cursor?
Cursor reads MCP servers from ~/.cursor/mcp.json (every project) or .cursor/mcp.json (one project). Add a "done" entry under mcpServers with "url": "https://trydone.app/mcp", then pick one of two ways to authenticate:
- Sign in with OAuth: open Customize in Cursor's sidebar, find done among your MCP servers, sign in when Cursor asks, and approve access in Done in the browser. Make sure done is switched on there.
- Use a token: generate an MCP token in Done under Settings → Agent access, export it as DONE_MCP_TOKEN, and add a headers entry with "Authorization": "Bearer ${env:DONE_MCP_TOKEN}". Reference the variable rather than pasting the token, because a project's .cursor/mcp.json is easy to commit by accident.
What can Cursor's agent do with my Done tasks?
Cursor gets the same tools any MCP client gets. next_task atomically picks and claims the highest-priority task that is ready, so two agents never take the same one. get_task loads the notes, subtasks, blockers and activity. comment_on_task posts progress where you'll read it, beside the task. When the work is done, complete_task closes a task marked auto-OK, and propose_completion posts a summary and parks every other task in pending approval.
Can Cursor's agent close a task without my approval?
Only if you marked that task auto-OK. Whether a task may close without a human is your grant, not the agent's: an agent can't set agent_ok when it creates or edits a task, and a task that isn't agent_ok can only reach done through your approval. Cursor's agent can do the work and propose it. It can't approve itself.
Should I use the global or the project mcp.json?
Use the global file if you want Done in every project. A project's .cursor/mcp.json is shared with anyone who opens the project, so keep any token out of it and use the ${env:DONE_MCP_TOKEN} reference. A token is per account, never shared, and can be read-only or limited to one project, so an agent working in one repo needn't see the rest of your board. Done doesn't launch Cursor or run it on a schedule; whatever starts the agent is what makes it check the queue.
Why won't Done connect in Cursor?
- Open the Output panel (Cmd+Shift+U, or Ctrl+Shift+U), choose MCP Logs, and look for the error from done.
- The desktop app receives the sign-in on localhost port 8787. If a local dev server is using that port, stop it while you sign in, or use the token setup.
- From Cursor's web or background agents the sign-in returns to www.cursor.com, which Done doesn't vouch for, so Done shows an unverified-app screen. Continue if you started the sign-in yourself.
Frequently asked questions
How do I add an MCP server to Cursor?
Add an entry under mcpServers in ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one project) with the server's url. For Done the URL is https://trydone.app/mcp. Then sign in from Customize in Cursor's sidebar, or send a token in a headers entry.
Is it safe to put a Done token in .cursor/mcp.json?
Not directly. A project file is easy to commit and is shared with anyone who opens the project. Reference an environment variable instead, "Bearer ${env:DONE_MCP_TOKEN}", and keep the token itself out of the file.
Why does the Cursor sign-in fail in the desktop app?
Cursor's desktop app receives the sign-in on localhost port 8787. If a local dev server is using that port, stop it while you sign in, or use the token setup.