OpenRig Released, Supports Saving and Restoring Claude Code and Codex Sessions

OpenRig, developed by Mike Schwarz, has been released. This tool is a Multi-agent Harness that defines Claude Code and Codex sessions as YAML files to configure them as a single team, enabling saving and restoration.

Key Features and Concepts

• Session Restoration and Persistence: Previously, restoring agent configurations upon reboot took about 30 minutes, and agents that were forgotten were lost. OpenRig solves this problem by saving snapshots with the rig down --snapshot command and restoring them with rig up.
• Seats and Pods: Maintains roles and cumulative context through fixed addresses called 'Seats', such as dev-owner@first-project. Teams are structured by grouping related Seats into 'Pods' that share common guidelines.
• Work Queue: Stores work items as DB rows to prevent loss during context compression or session replacement. Manages workflows via transactional processing using the rig queue create/claim/handoff commands.

Tech Stack and Environment

• Components: Composed of a local daemon, CLI, Terminal UI (TUI), and MCP server, written in TypeScript.
• Execution Environment: Supports macOS and Linux (no native Windows support), and requires Node.js 20/22/24 and tmux.
• Deployment: Can be installed via the npm package @openrig/cli, with no separate hosting costs; only the costs of each user's logged-in Claude/Codex account apply.

Precautions

• Environment Changes: Running rig setup may modify existing configuration files such as ~/.tmux.conf, ~/.claude.json, and ~/.codex/config.toml, so prior review is necessary.
• Permission Management: The default YOLO mode is disabled; during managed execution, Claude uses --permission-mode acceptEdits and Codex uses -s workspace-write mode.