Back to browse
GitHub Repository

Self-evolving AI agent framework with 5-layer safety gatekeeper. Agents observe failures, propose fixes, and safely apply them. Built on HKUDS/nanobot.

38 starsPython

Geneclaw – An AI agent framework that safely evolves its own code

by geneclawai·Feb 18, 2026·1 point·0 comments

AI Analysis

PassBold BetNiche Gem
The Take

This actually tries something non-trivial: a Geneclaw Evolution Protocol (GEP) that turns failures into structured evolution proposals, complete with unified diffs, risk levels and rollback plans. The five-layer gatekeeper + git-branch apply + automated test rollback combo shows real operational thinking — and the default dry-run mode is the correct safety posture. It's early and specialist: adoption will hinge on more demos, audits and battle-tested examples of autonomous fixes gone right (or safely stopped).

Category
Target Audience

ML engineers, SREs and developers building autonomous agents, security-conscious dev teams exploring safe self-repair

Similar Projects

AI/ML●●Solid

Putting Git on AI Agents

Git for agent cognition—clever framework, but no working implementation yet.

Big BrainWizardry
vichoiglesias
223mo ago