multi-agent-shogun is a system that supports running multiple AI coding CLI instances simultaneously, orchestrating them like a feudal Japanese army. It supports Claude Code, OpenAI Codex, GitHub Copilot, and Kimi Code. It aims to solve the problem of high API costs associated with running multiple agents by using a zero-overhead YAML-based hierarchy and tmux. The core technology involves a hierarchical architecture with a Shogun orchestrator, a Karo task distributor, and Ashigaru worker agents.
The project distinguishes itself through its zero-overhead architecture, utilizing YAML files for coordination instead of API calls. Its transparent design ensures complete visibility into agent workflows via live tmux panes and easily readable YAML files. The bottom-up skill discovery mechanism allows agents to automatically identify and propose reusable patterns, enhancing long-term productivity. Support for multiple CLIs (Claude Code, Codex, Copilot, and Kimi) allows flexibility and cost optimization.
- Parallel Execution: Enables concurrent execution of multiple AI coding agents, significantly reducing task completion time.
- Multi-CLI Support: Supports Claude Code, OpenAI Codex, GitHub Copilot, and Kimi Code CLIs, providing flexibility and vendor choice.
- Zero Coordination Overhead: Uses YAML files for agent communication, minimizing API calls and associated costs.
- Real-time Observability: Provides a live dashboard and visible tmux panes for monitoring agent progress and decision-making.
- Automatic Skill Discovery: Identifies reusable patterns from agent workflows, automatically creating new commands and improving long-term efficiency.
- Flexible Configuration: Uses YAML configuration files for easy customization of agent behavior and workflows.
- Extensible Architecture: Designed with a modular architecture for easy integration of new CLIs and agent types.
multi-agent-shogun is an active project with recent commits and ongoing development. The documentation is comprehensive, and the community is growing. While not entirely production-ready, it demonstrates a robust architecture and core functionality. The system shows good stability for experimentation and offers a promising approach to multi-agent workflows.
multi-agent-shogun benefits developers by offering a cost-effective and transparent way to leverage multiple AI coding agents in parallel. It is ideal for automating coding tasks, exploring different AI models, and improving developer productivity. Unlike traditional multi-agent frameworks that rely on API calls, Shogun's architecture provides significant cost savings and increased control over AI workflows.