This project orchestrates multiple Claude Code AI agents in parallel to systematically improve codebases, targeting developers and teams seeking automated bug fixing and best practice implementation. It offers significant benefits by leveraging distributed AI power for large-scale code maintenance and enhancement.

How It Works

The framework utilizes a Python script to manage multiple claude sessions within tmux panes, acting as independent agents. These agents coordinate through a sophisticated lock-based system for conflict prevention, allowing them to work on different parts of the codebase simultaneously. The system monitors agent health, context usage, and progress, with features for auto-recovery and detailed reporting.

Quick Start & Requirements

• Install: Clone the repository and run ./setup.sh.
• Prerequisites: Python 3.13+, tmux, claude command configured, git, project-specific tools (e.g., bun, mypy), uv, and optionally direnv.
• Verification: Run claude-code-agent-farm doctor --path /path/to/project.
• Docs: README

Highlighted Details

• Supports 34 technology stacks with dedicated configurations and best practices guides.
• Features an advanced "Cooperating Agents Workflow" for complex, coordinated improvements using a distributed lock and work registry system.
• Includes 24 integrated tool setup scripts for various development environments (Python, Go, Rust, Next.js, etc.).
• Offers a real-time monitoring dashboard within tmux showing agent status, context usage, and errors.

Maintenance & Community

Created by Jeffrey Emanuel. Contributions are welcome via pull requests.

Licensing & Compatibility

MIT License. Compatible with commercial and closed-source projects.

Limitations & Caveats

Requires a specific cc alias setup for claude with elevated permissions (ENABLE_BACKGROUND_TASKS=1 claude --dangerously-skip-permissions). The effectiveness of the "Cooperating Agents Workflow" is heavily dependent on the LLM's understanding of the prompt and coordination protocol.