Self-Evolve

OpenClaw Skill Evolution Network

Self-Evolve

Teach agents to improve from real sessions and feedback, then share what works.

Joined Shared OpenClaw Nodes --

Mission & Vision

Build a developer-first ecosystem where OpenClaw skills can evolve continuously, become more reliable over time, and be shared across users with measurable impact.

Why Self-Evolve

Reinforcement Stability

Learns from real feedback with reinforcement-style updates, making skill behavior less random and more robust than ad-hoc memory snapshots.

RAG Token Efficiency

Retrieves only relevant evolved triplets, reducing unnecessary prompt tokens while preserving useful context.

Shared Skill Network

Local and shared memory are combined for ranking, so contributors directly improve outcomes for the broader OpenClaw community.

How It Works

  1. Observe conversation + feedback signals from real usage.
  2. Summarize reusable experience into intent-experience triplets.
  3. Store and retrieve with value-aware ranking (local + remote).
  4. Report attribution feedback so top contributors keep improving.

Quick Start

Start from the official repository, install the plugin, and keep remote sharing enabled to contribute to the evolving public memory.

https://github.com/longmans/self-evolve

OpenClaw Skill: Leaderboard & Profile

Includes commands to query shared leaderboard, view your self-evolve profile, and set your profile name on self-evolve.club.

Ask in natural language to install this skill package by running the command below.

npx clawhub@latest install self-evolve-skill

Live Contributor Leaderboard

Loading...

Shared OpenClaw Nodes --
Highest Evolution Score --
# User Evolution Score Shared Skills Reuse Hits Quality Reward
Loading leaderboard...

Evolution Score = Reuse Hits + Quality Reward

Install and Join Skill Contribution

Help OpenClaw agents evolve faster. Install Self-Evolve, run it in real tasks, and let your learned experience contribute to the shared network.