This project provides a REST API and associated tools for coordinating multiple LLM agents in software development projects. It addresses the challenge of managing complex, multi-agent workflows by offering features like task hierarchy, Git integration, and document-based communication. The system is designed for developers and teams leveraging LLM agents for coding tasks, aiming to streamline project execution and enhance productivity.

How It Works

The system utilizes a FastAPI backend with a SQLModel ORM for data management, supporting SQLite and MySQL. Agent coordination is achieved through document-driven messaging and a polling-based update mechanism. Key architectural choices include file-based agent instructions and a stateless agent design, promoting flexibility and scalability. A service registry with heartbeat monitoring provides system health insights, while Git branch integration automates workflow decisions based on task complexity.

Quick Start & Requirements

• Primary install/run: Clone the repository, run ./setup/universal_setup.sh for platform-specific environment setup, and then ./start.sh to launch the server.
• Prerequisites: Python, Git. Specific pydantic versions are required for x86_64 environments (e.g., Claude Code).
• Links: Documentation is available within the docs/ directory, including CLAUDE.md for Claude Code specifics and SAMPLE_AGENT_WORKFLOW.md.

Highlighted Details

• Task Management: Supports an Epic → Feature → Task hierarchy with role-based assignment and skill levels.
• Git Integration: Automates Git workflows based on task complexity (e.g., major tasks requiring PRs, minor tasks allowing direct commits).
• Agent Communication: Features document-based messaging with @mention support and a notification system.
• Monitoring: Offers real-time CLI and Web dashboards for project overview, task status, agent activity, and service health.
• Claude Code Integration: Includes an MCP server for natural language commands and specific setup scripts for Claude Code environments.

Maintenance & Community

The project is attributed to Timo Railo, 2025. No specific community channels (like Discord or Slack) or explicit roadmap links are provided in the README.

Licensing & Compatibility

The project is released under the MIT License, allowing for broad use and modification. Compatibility notes highlight the need for specific pydantic versions when running in x86_64 environments, such as those used by Claude Code.

Limitations & Caveats

The API may exhibit a 3-minute response delay when no tasks are available. Database migrations might require manual table drops and recreation in the current version. Some user feedback indicates potential setup or functionality issues ("Doesn't work for me.").