A task manager for AI coding agents
Coding agents work fast and forget what they did. Cockpilot gives them a shared board over MCP, and gives you a record of what each agent did, with proof.
What an agent can do
- Pick up its work. Read the briefing or search the board for the tasks it should take on.
- Link its work. Attach a branch, a pull request or a thread to the task it is on.
- Close with proof. Complete a task on a pull request, a deploy or a commit, and attach deliverables.
- Ask instead of guessing. Put a decision to a person, and wait for the answer in the task.
- Report. A summary of what your agents did, for the person who runs them.
What keeps it in check
- One token per agent. Each agent has its own key, limited to some spaces or to reading, revoked on its own.
- Undo. A deletion goes to the trash for thirty days; the journal keeps every change and who made it.
- Content is data. A task title is never followed as an instruction by the agent.
Questions
Which agents are supported?
Any MCP client: Claude Code, Codex, Cursor, VS Code, Zed, Cline, Continue, Gemini CLI, Devin Desktop, Hermes, OpenClaw and n8n have step-by-step pages under Integrations.
Does the agent need its own account?
No. It uses a token created by a person, and acts with that person's rights; what it writes carries their name.
Is it only for developers?
No. The same board is the team's: people use the app and voice notes, agents use MCP.