Player not loading? Watch on YouTube
OpenRig is an open source CLI for coordinating coding agents in groups called rigs. The speaker demonstrates how agents message one another and control each other's terminals through the rig command. The workflow combines Claude Code and Codex, with a browser UI that displays the agent topology and opens individual CLI sessions. OpenRig is designed to work with tmux.
The walkthrough explains a setup with two orchestrators, one Claude and one Codex. The speaker says these can help restore context when either session resets or compacts, accept permission prompts for other agents and nudge agents that stop early. They describe keeping orchestrator sessions running for weeks; this is their reported experience, rather than a guarantee of unattended operation.
YAML specs define an agent's role, including skills, starting files, sub-agents and hooks. The UI browses a folder of specs and lets users launch them. A discovery tab can scan for agents already running on the machine and add them to a rig.
The speaker describes the project as raw and says the UI is not fully functional, recommending the CLI as the main interface. They also say regular Claude Code and Codex subscriptions can support these sessions without separate API costs. The demonstration concerns agent coordination, not how to run models locally.
