Player not loading? Watch on YouTube
OpenRig is an open source project for saving and managing teams of terminal-based coding agents. Mike, its creator, describes a personal setup with more than 70 Claude and Codex agents on a single Mac mini. Most are idle during the introduction, retaining chat history and context about the code they worked on. The count describes his setup, not a hardware requirement or performance benchmark.
The video explains how his workflow grew from Cursor into multiple Claude Code sessions, then added Codex CLI. Agents first exchanged markdown files; later, he used tmux for direct messaging. OpenRig aims to preserve the team's arrangement: which agents exist, their workspaces, and how they communicate.
Agents use a CLI and control plane, while the operator uses a UI to inspect the fleet and reach individual terminals. Work stays in those terminals. Built-in queues and handoffs help agents pass context and request human judgment.
Mike calls groups of related agent contexts "context domains" and argues that they can help preserve shared knowledge through context compaction. His discussion of combined context capacity is a conceptual argument, not a measured result. This is an overview of AI agent coordination rather than a local LLM setup guide. He describes OpenRig as early and says his own configuration still needs maintenance.
