OpenRig

OpenRig coordinates coding agents for developers working on projects that last weeks or months. You talk to a lead agent, which delegates to specialists and brings back results or decisions that need your attention.

• Persistent teams: Claude Code, Codex and Pi can work together with continuing roles and retained context, while specialists use temporary subagents for smaller jobs.
• Owned assignments: Agents assign work to named teammates and track who has claimed it, with ownership and progress available for later follow-up.
• Direct communication: Specialists can ask teammates about earlier decisions without creating a tracked task for each question.
• Team overview: A terminal interface shows agent organization, current work and requests for attention; you can also inspect or work directly in each agent's terminal session.
• Session continuity: OpenRig can adopt existing Claude Code and Codex sessions, save and restore team arrangements, and add or remove agents while a team is running.

The coordination layer runs on your own macOS or Linux machine or server. Windows support uses WSL2; native Windows isn't supported. OpenRig is open source under Apache 2.0 and requires Node.js and tmux. Its MCP server lets agents communicate and manage the team themselves.

Self-hosting applies to coordination. The selected providers handle model usage, and their usage costs still apply. The Codex starter requires a Codex login.