Summary

This project provides a clean-room reimplementation of the "Claude Code" agent harness, focusing on harness engineering, tool orchestration, and runtime context management. It targets developers and researchers interested in understanding and building advanced AI agent systems, offering a faster, memory-safe Rust-based runtime and an AI-assisted development workflow. The primary benefit is a functional, open-source exploration of complex agent architectures without relying on proprietary code.

How It Works

The project is undergoing a significant transition to a Rust implementation for a faster, memory-safe harness runtime, intended to be the definitive version. Development is heavily augmented by oh-my-codex (OmX), a workflow layer built on OpenAI's Codex, utilizing modes like $team for parallel code review and $ralph for persistent execution and verification. This AI-driven approach orchestrates the entire porting process, from understanding harness structures to generating a working Python codebase with tests, capturing architectural patterns without direct code copying.

Quick Start & Requirements

• Primary install/run: Python 3.x environment.
• Commands:
  • Render Python porting summary: python3 -m src.main summary
  • Print workspace manifest: python3 -m src.main manifest
  • List modules: python3 -m src.main subsystems --limit 16
  • Run verification: python3 -m unittest discover -s tests -v
• Prerequisites: Rust toolchain (for the ongoing Rust port), Python 3.x. Specific hardware or CUDA requirements are not detailed.
• Links: Project repository: https://github.com/instructkr/claw-code

Highlighted Details

• Achieved over 50K GitHub stars in just 2 hours, positioning itself as a historically fast-growing repository.
• Development is significantly accelerated and guided by oh-my-codex (OmX) and OpenAI's Codex, enabling parallel review and persistent execution loops.
• Focuses on "harness engineering"—the study of how AI agent systems wire tools, orchestrate tasks, and manage runtime context.
• The project aims to be a definitive version, with an active Rust port targeting enhanced performance and memory safety.

Maintenance & Community

The project is actively developed in collaboration with the creator of oh-my-codex. Continued research and development are supported by GitHub sponsorships for the primary maintainer. A community Discord server is available for discussions on LLMs, harness engineering, and agent workflows.

Licensing & Compatibility

The repository does not explicitly state a software license. It emphasizes a "clean-room Python rewrite" and discusses the legal and ethical implications of AI code reimplementation, suggesting an intent to avoid direct use of proprietary code. Compatibility for commercial use or linking with closed-source projects is undetermined due to the lack of a specified license.

Limitations & Caveats

The current Python workspace is not yet a complete runtime-equivalent replacement for the original system, containing fewer executable runtime slices. The repository does not claim ownership of the original "Claude Code" source material and is not affiliated with Anthropic. The rapid star growth is partly attributed to the project's association with leaked "Claude Code," which may carry associated risks.