mvschwarz/openrig

OpenRig is a multi-agent harness that orchestrates Claude Code and Codex together as a single coordinated system.

The tool addresses the fragmentation of running multiple AI coding agents by replacing scattered terminal sessions with a persistent, organized team managed through a unified interface. It works by defining agent teams in YAML configuration, then launching them as a coordinated group where a lead agent can delegate work to specialists and surface results requiring human review. The system maintains persistent context and workspace state across agent interactions, allowing teams to build on prior work without losing continuity.

OpenRig suits developers who want to run multiple AI coding agents on the same codebase without managing separate sessions manually. It works best for projects where you want a lead agent to coordinate specialists—for instance, having one agent propose changes while another reviews them—and where you need to preserve the team's work and context at consistent addresses. The tool requires Node.js 22 or 24 and tmux on macOS or Linux; Windows is not supported natively, and WSL2 has not been tested. Before launching, you should review what OpenRig changes on your machine, including provider hooks and workspace trust settings, and back up relevant files. The setup process includes a plan review step and requires authenticated Codex for the starter configuration.

Development activity shows consistent, focused iteration on core functionality. The project maintains clear documentation including a guided first-use path and compatibility notes for different Node.js versions. Updates and demonstrations are shared regularly through the maintainer's public channel. The tool includes a terminal user interface for monitoring agent state, runtime, model selection, and context across team members, with separate views for operational support and shared dashboards.