Clawland-AI/Geneclaw
Self-evolving AI agent framework with 5-layer safety gatekeeper. Agents observe failures, propose fixes, and safely apply them. Built on HKUDS/nanobot.
The framework implements a closed-loop self-improvement cycle via the **Geneclaw Evolution Protocol (GEP v0)**: agents record all interactions as JSONL events, diagnose failures heuristically or with LLM assistance, generate structured evolution proposals with unified diffs, and route them through a 5-layer gatekeeper (allowlist, denylist, diff size, secret scan, code patterns) before Git-branched application with automated test execution. Built atop nanobot's lightweight agent-loop and tool system, it provides observability through append-only event stores, dry-run-by-default safety, Streamlit audit dashboards, and configurable autopilot loops for risk-based auto-approval.
Stars
22
Forks
4
Language
Python
License
MIT
Category
Last pushed
Feb 18, 2026
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/agents/Clawland-AI/Geneclaw"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
arielshad/balagan-agent
Chaos Engineering for AI Agents
evoplex/evoplex
Evoplex is a fast, robust and extensible platform for developing agent-based models and...
MrTsepa/autoevolve
AI agent evolving strategies through automated self-play overnight. Generic framework with...
selinayfilizp/decision
The missing layer in the AI agent stack. Teach your agent how you decide. Generates a portable...
SharedIntellect/quorum
Evidence-grounded quality validation for AI agent outputs. Rubric-driven, multi-critic, open source.