Multi-Agent Shogun

Multi-Agent Shogun runs and coordinates separate AI coding assistants on your computer or a Linux server. It's for developers dividing implementation and research across parallel tasks. Coordination happens locally, but Claude Code’s standard cloud connection sends prompts and model outputs over the network to their model provider; other agents use their chosen CLI's model access and authentication.

The system supports Claude Code, OpenAI Codex, GitHub Copilot, Kimi Code, OpenCode, Cursor and Antigravity. Different agents can use different CLIs. A Shogun receives requests, a Karo distributes work to Ashigaru workers, and a Gunshi takes on analysis, evaluation and design. You can submit further requests while work continues.

Each agent has a visible tmux pane, and a dashboard tracks team progress. Instructions, decisions and reports live in readable YAML files that you can compare or keep in version control. File-based messaging gives you an inspectable record of assignments and handoffs.

Multiple projects can share the same coordinator while retaining separate project context. Memory MCP preserves preferences across sessions, and workers can propose recurring patterns as reusable skills for your approval. Task dependencies let the manager hold work until prerequisite tasks finish.

The project is open source under the MIT license and runs on Linux and macOS, or Windows through WSL2. An Android companion app provides SSH access to agent sessions and the dashboard; optional ntfy integration carries phone commands and progress notifications.