A universal AI agent framework, Gitclaw treats agents as Git repositories, versioning their identity, rules, memory, tools, and skills. This approach targets developers and power users seeking robust agent management, enabling granular control, history tracking, and collaborative development through familiar Git workflows. The primary benefit is managing AI agents with the same rigor and tooling applied to software development.

How It Works

Gitclaw's core innovation is its "git-native" architecture. Agent configurations, including personality (SOUL.md), behavior (RULES.md), runtime settings (agent.yaml), memory, tools, and skills, are stored as files within a Git repository. This allows agents to be forked, branched, and their evolution tracked via git log and git diff. The framework leverages declarative YAML for defining tools and skills, alongside programmatic interfaces via an SDK, offering flexibility in agent implementation.

Quick Start & Requirements

• Primary Install: A one-command installer is provided: bash <(curl -fsSL "https://raw.githubusercontent.com/open-gitagent/gitclaw/main/install.sh?$(date +%s)"). Alternatively, npm install -g gitclaw.
• Prerequisites: Node.js 18+, npm, and git are required.
• Setup: The installer guides through API key setup and launches a local voice UI. Basic agent execution requires setting an API key (e.g., export OPENAI_API_KEY="sk-...") and running gitclaw --dir ~/my-project "Your prompt".
• Links: Install script: https://raw.githubusercontent.com/open-gitagent/gitclaw/main/install.sh.

Highlighted Details

• Git-Native Configuration: Agent identity, rules, memory, tools, and skills are version-controlled files within a Git repository.
• Declarative & Programmatic Tools: Tools can be defined via YAML files or programmatically through the TypeScript SDK.
• Extensibility: Supports composable skills, lifecycle hooks (script or programmatic), and a plugin system for adding custom functionality.
• Multi-Model Support: Integrates with various LLM providers (Anthropic, OpenAI, Google, etc.) via the pi-ai backend, allowing fallback models.
• Local Repo Mode: Enables agents to clone GitHub repositories, work on them, and auto-commit changes to a session branch.
• SDK: Provides an in-process TypeScript SDK mirroring the Claude Agent SDK pattern for programmatic agent control and tool/hook registration.

Maintenance & Community

No specific details regarding maintainers, sponsorships, or community channels (like Discord/Slack) are present in the provided README.

Licensing & Compatibility

The project is licensed under the MIT License. This license generally permits commercial use and integration into closed-source projects.

Limitations & Caveats

The README does not explicitly detail limitations. However, the core functionality is tied to Git repositories, and the sandbox mode is mutually exclusive with the local repo mode. Reliance on external LLM providers necessitates their availability and API key management.