mvschwarz

openrig: Coordinate Claude Code and Codex Agent Teams

OpenRig uses Claude Code and Codex, coordinated through tmux, to run persistent agent teams from YAML-defined roles and shared context. It is for developers who want to manage coding agents as an organized, recoverable system on macOS, Linux, or WSL2.

ActiveTypeScriptApache-2.0AI AgentsCLIAgent Orchestration
Stars
8.9k
Forks
776
Last commit
today
Contributors
98
Releases, 12 months
71

Our take

Activeregular commits and releases

Health score

77/100

How it is scored
Activity
87
Community
61
Issues
74
Pull requests
85
  • Merges more pull requests than 98% of the projects we track
  • Releases more often than 90% of the projects we track
  • More contributors than 77% of the projects we track

OpenRig is an actively developed, feature-rich way to turn Claude Code and Codex sessions into persistent teams, though its rapid evolution and concentrated contribution base call for deliberate evaluation before relying on it for critical workflows.

Good fit if

  • You use Claude Code or Codex and want both providers managed in the same team through tmux.
  • You need persistent roles, coordination, and recovery rather than a one-off multi-agent script.
  • You can accommodate the supported Node.js and tmux setup, and want a project with frequent releases and active issue and pull-request handling.

Look elsewhere if

  • You need native Windows support rather than WSL2, or cannot run tmux.
  • Your workflow depends on a hosted, fully managed service instead of a local daemon and CLI.
  • You need a mature project with contributions spread broadly across its maintainers; most measured contributions come from one person.
All health signals
Last commit2026-11-22 (today)
Commits, last 90 days100+
Releases, last 12 months71 (latest v0.7.21, 2026-11-21)
Contributors98 (top contributor: 75% of commits)
Issues closed, last 90 days13+ (typically closed in 1 day)
Pull requests merged, last 90 days97 (typically merged in 0 days)
Project age8 months

Checked on 2026-11-22 with the GitHub API.

Overview

OpenRig is a local multi-agent harness built around Claude Code and Codex, with tmux managing their terminal sessions. Define a team in YAML, then use one system to start, inspect, coordinate, and recover its agents instead of juggling separate terminal sessions.

It is aimed at developers who want coding agents to work together over ongoing tasks, with stable role-based addresses, shared context, and a queue for owned work. OpenRig also supports terminal nodes and Pi and Oh My Pi through RPC runners, but Claude Code and Codex are the central providers described by the project. It runs on macOS and Linux, with WSL2 as the supported Windows route.

Key Features

  • Define agent teams as YAML RigSpecs, with pods, roles, communication edges, and continuity policies.
  • Start teams with rig up, and manage their tmux sessions through the CLI, TUI, and MCP server.
  • Inspect topology, agent state, feeds, projects, and system health in the terminal UI.
  • Send messages, broadcast to teams, use chatrooms, and track assigned work in a queue.
  • Snapshot a team and restore it by name, with per-agent outcomes reported during recovery.
  • Discover existing Claude Code and Codex sessions in tmux and adopt them into managed teams.
  • Install reusable rig bundles and use built-in team templates, including starter, workshop, and factory.

Use Cases

  • A developer wants a builder and reviewer to collaborate on one bounded repository change, using the starter team.
  • A small product team needs a lead, builder, QA agent, and reviewer to coordinate ongoing work in one repository.
  • An OpenRig operator wants to preserve team structure and recover agent sessions after stopping or restarting a rig.
  • A team already running agents in tmux wants to bring those sessions under OpenRig management rather than recreate them.

What you need

Detected in the repository

  • Node.js ^22.14.0 || ^24 (from package.json)
  • A test suite and automated checks on GitHub Actions

License in plain words

Apache-2.0permissive

  • Commercial use: yes
  • Modify and redistribute: yes
  • You must keep: the license, the NOTICE file and a note of your changes
  • Share your changes: no
  • Includes an explicit patent grant from the contributors:

A summary, not legal advice: the LICENSE file is what applies.

Getting Started

Install the CLI with Node.js 22 or 24:

npm install -g @openrig/cli
rig setup

OpenRig also requires tmux and Claude Code or Codex signed in. See the README and getting started guide for setup details and machine changes.

Alternatives

  • ruflo: Ruflo adds persistent memory, workflow automation, and MCP integrations, while OpenRig centers on YAML-defined teams coordinated through tmux.
  • oh-my-claudecode: oh-my-claudecode focuses on Claude Code delegation and verification, while OpenRig coordinates both Claude Code and Codex in persistent teams.
  • rimz: RimZ manages agents in existing tmux or Zellij sessions with status and messaging tools, while OpenRig defines teams and shared context in YAML.
ProjectLanguageLicenseStarsStatus
openrigTypeScriptApache-2.08.9kActive
rufloTypeScriptMIT73.8kActive
oh-my-claudecodeTypeScriptMIT39.6kActive
rimzRustMIT33New

Considerations

  • OpenRig is young and changing quickly: the project reports frequent commits and releases, while most measured code contributions come from one contributor. Teams should evaluate upgrades and recovery behavior in their own environment.
  • Setup and operation write instance state and can modify tmux and provider configuration, including trust settings and executable hooks. The README advises backing up relevant files and notes that preservation and rollback are not guaranteed.
  • Node.js 22 or 24 and tmux are required. The supported operating systems are macOS and Linux; Windows use is through WSL2, and the project notes its automated tests do not run on WSL2 yet.
  • Some bundled rigs require Docker for service-backed workloads. Docker is not a general requirement for OpenRig itself.

Found this useful?

Share it with someone who would like openrig.