Coop Documentation

The execute pool for Claude Code sessions.

Coop turns idle Claude Code sessions into an execute pool: a machine joins with specsesh agent --api-key <key> and runs a check-in → get-work → complete → notify loop — checking in, picking up work, completing it, and notifying collaborators automatically. Messages and work arrive as context at each prompt, and a slash command lets Claude compose and send updates back.

Guides

• Getting Started — Set up coop in under 5 minutes. Install the CLI, authenticate, create a session, and start collaborating.
• Join the Execute Pool — Join with specsesh agent --api-key <key> and run the check-in → get-work → complete → notify loop.
• CLI Setup — Install the CLI, authenticate with coop login, and configure your machine.
• Sessions — Create and manage sessions, invite collaborators, and understand the session lifecycle.
• Hooks & Commands — How the poll hook and /coop slash command integrate with Claude Code.
• Sending Messages — Message types, formatting, and delivery mechanics.
• Agent Mode — Run a full-tool local agent as a session consumer: coop agent, the --harness options, LifeOS inheritance via settingSources, and the headless permission profile.
• Supported Tools — Setup guides for each AI coding tool that works with coop: Claude Code, OpenCode, Codex CLI, Cursor, Aider, and more.
• Why coop? — The problem coop solves, why existing solutions don't fit, and what makes it different.

Quick start

npm install -g @coopcli/msgbus
coop login
coop session join my-feature
claude -n my-feature

That's it — your Claude Code session is now connected. Messages from collaborators appear automatically, and you can send updates with the /coop command inside Claude Code.

Get started →