SkillClaw evolves AI agent skills from real interactions, enabling collective evolution across sessions, agents, and devices. It consists of a local client proxy and an optional evolve server, with broad Hermes/OpenClaw/OpenAI-compatible integrations and a path for single-user or team usage.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
SkillClaw is a system for evolving reusable AI agent skills from real session data. It provides a client proxy that intercepts agent requests and records session artifacts, and an optional evolve_server service that reads session data from shared storage, evolves or creates skills, and writes them back. It supports two engines: workflow (Summarize → Aggregate → Execute) and agent (OpenClaw-driven workspace with direct skill editing). The components share a storage layer and a SKILL.md format. It targets single-user local usage or team usage with shared storage for group evolution.
How it works
The system comprises two components: a Client Proxy (local API proxy) and an Evolve Server (optional). Both read/write to shared storage (Alibaba OSS / S3 / local filesystem) and use the SKILL.md format. The client exposes endpoints such as /v1/chat/completions and /v1/messages via a local proxy. The evolve_server can run on the same machine or elsewhere to automatically refine skills in the background and supports two engines: workflow and agent. Prerequisites include Python >= 3.10 and OpenAI-compatible APIs or AWS Bedrock. The project emphasizes deduplication, aggregation, and cross-session skill improvement, enabling collective evolution across multiple agents, devices, or users.
Getting started
Prerequisites:
- macOS, Linux, or Windows
- Python >= 3.10
- An OpenAI-compatible API provider or AWS Bedrock
- Install openclaw only if you intend to use the openclaw CLI integration or the server agent engine
Path A (single machine):
- Install SkillClaw from this repository. If you already have the source checkout, skip
git clone. macOS / Linux:
git clone https://github.com/AMAP-ML/SkillClaw.git && cd SkillClaw
bash scripts/install_skillclaw.sh
source .venv/bin/activate
Windows PowerShell (manual install):
git clone https://github.com/AMAP-ML/SkillClaw.git
Set-Location SkillClaw
python -m venv .venv
.
.venv\Scripts\Activate.ps1
python -m pip install -U pip
python -m pip install -e ".[evolve,sharing,server]"
- Generate a local config.
skillclaw setup
- Start the client proxy and verify health.
skillclaw start --daemon
skillclaw status
PROXY_PORT="$(skillclaw config proxy.port | awk '{print $2}')"
curl "http://127.0.0.1:${PROXY_PORT}/healthz"
Path B (join a shared group) and other config examples are provided for sharing via OSS/S3 or local storage and for enabling Hermes integration.
Recent releases
Latest release entries: none.
Traction
Stars: 2351 Forks: 234 Open issues: 42
Behind the repo
License: MIT Language: Python Created: 2026-04-10 Last push: 2026-06-02 Topics include: agent, agentic-ai, ai-agent, collective-intelligence, continual-learning, hermes, llm, llms, openclaw, self-evolving, skill-evolution, skill-learning, skills
Caveats
Prerequisites specify Python >= 3.10 and an OpenAI-compatible API or AWS Bedrock. The repository notes that its latest releases include documentation updates and integrations but shows no releases declared in RELEASES (latest 0).