Skip to content
← Tools / Skills
ToolUltraworkers/10 trustVerified

Claw Code

Open-source AI coding agent framework — clean-room Python/Rust harness, ~190K GitHub stars.

Visit official site →

About

Claw Code is an open-source AI coding agent framework built in Python and Rust, launched publicly in April 2026 as an independent, auditable foundation for AI-assisted development. It targets the "agent harness" layer — the control plane connecting models to tools, file systems, and task workflows — which has remained largely proprietary. Built as a clean-room rewrite, it gives developers a fully inspectable and extensible harness architecture. The repository became the fastest in GitHub history to 100K stars, sitting near 190K+ stars with an MIT license. The repo includes a Python workspace, test suite, and CLI utilities for summaries, manifests, and subsystem inspection, with a Rust port in progress for a faster production runtime.

The honest picture: Claw Code's origin story is tied to the March 2026 Claude Code source leak, which raises uncomfortable questions even for a clean-room rewrite. Star velocity reflects hype as much as production readiness, and the Rust port is still in progress. As a harness rather than a full agent product, it requires significant assembly before it does useful work — it's infrastructure for builders, not a tool for end users.

What it can do

  • Open, auditable agent harness architecture (Python + Rust)
  • Tool orchestration and agent runtime primitives
  • CLI utilities for summaries, manifests, and parity audits
  • Clean-room implementation with no proprietary weights
  • Extensible subsystem design for custom agent builds

Setup

Clone github.com/ultraworkers/claw-code, install Python dependencies, and run the CLI utilities. Connect your preferred model provider via configuration.

Prerequisites

Python 3.x; a model API key; familiarity with agent harness concepts.

Use cases

ai-agentsopen-sourcedeveloper-tools

Works with

Trust score — 9/10

/10 trustVerified

Scored 9/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 54; Trending score 90/100 from researcher assessment.

★ 190,000 GitHub stars

Sources

Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.