Multi-Agent Shogun
An open-source AI coding agent orchestrator that runs on Linux, macOS and Windows via WSL2, with support for Claude Code, Codex and other coding CLIs.
Multi-Agent Shogun runs a team of AI coding assistants on your computer or server, with separate roles for delegation, task management, implementation and analysis. It's for developers who want parallel work across projects while keeping each agent's activity visible. The orchestration runs locally on Linux, macOS or Windows through WSL2; the connected coding services handle model access through their own authentication or provider API keys.
Its samurai-inspired hierarchy gives the Shogun responsibility for receiving requests, the Karo for dividing work and checking quality, and Ashigaru workers for implementation. A Gunshi strategist handles analysis, evaluation and design. You can submit another request while existing tasks continue in the background.
Agents coordinate through YAML files on disk and run in separate tmux panes. That makes instructions, decisions and reports available for inspection or version control, while a dashboard shows task progress. The file-based coordination doesn't require separate model API calls to pass messages between agents.
The system supports Claude Code, OpenAI Codex, GitHub Copilot, Kimi Code, OpenCode, Cursor and Antigravity. Shared instruction templates keep role rules consistent across those CLIs. Agents can also propose reusable skills based on patterns they encounter during work, with user approval before a proposal becomes a permanent skill.
The code is open source under the MIT license. An Android companion app provides remote monitoring and commands over SSH, screenshot transfers and voice input through Google Speech API.
