OpenRig

OpenRig is an open-source harness for running persistent teams of AI coding agents. Instead of chatting with a single assistant, you talk to one lead agent that delegates work to specialists, tracks what they are doing and brings the answers back to you. The project has earned about 8,000 stars on GitHub, is written in TypeScript under the Apache 2.0 license, and was still actively developed in 2026.

Overview

OpenRig lets Claude Code, Codex and Pi agents work together as a durable team on projects that stretch over weeks or months. Each agent keeps its own role, context and responsibilities, can spin up temporary subagents for smaller jobs, and stays reachable after the original conversation has moved on. The system is self-hosted and terminal-based: agents live in seats you can inspect through the OpenRig TUI, which shows how the team is organized, what everyone is working on and where your attention is needed. Work is routed with simple commands such as rig send to ask a specialist directly, or rig queue to assign a task to a named teammate who claims it and keeps its progress on record.

Key Features

• Persistent agent teams built from Claude Code, Codex and Pi, with roles and shared context
• Lead-agent workflow: one conversation in front, a whole specialist team behind it
• Direct messaging (rig send) and an owned work queue (rig queue) for long-running tasks
• Terminal UI showing the whole team’s organization, activity and open questions
• Agents can spawn their own temporary subagents for smaller sub-tasks
• Self-hosted and tmux-based, so teams keep running on your own machine

💳 Pricing

Free. OpenRig is open source under the Apache 2.0 license and runs on your own hardware; you only pay for the underlying agent subscriptions or API usage of the tools you plug into it.

Who Is It For

Developers and small teams running long-lived, multi-part projects who feel the limits of a single coding agent and want a coordinated crew of specialists with memory, ownership and visibility instead.