AIエージェント
evolver
A GEP-powered self-evolving engine for AI agents — auditable evolution with Genes, Capsules and Events.
★ 9.1k スターGPL-3.0
ガイドを見る →reefは、Human-Agent-Societyが公開しているオープンソースのAIエージェントプロジェクトです（GitHub: Human-Agent-Society/reef）。GitHubスターは7,504、フォークは698で、Apache-2.0ライセンスの下で公開されています。GitHubでの最終更新は2026年10月8日です。
Reef — infrastructure for continually self-improving agents: weight-training and harness-evolving deployments.
Choose your improvement path — weight training, harness evolution, or both — and identify which agent traffic and feedback signals will drive it.
Clone Human-Agent-Society/reef from GitHub (or use the install path in the run steps below) and skim its README for the parts relevant to your goal.
Install reef-infra from PyPI for use, or clone and install from source for development and the full recipes.
Stand up a minimal deployment, send a provider-compatible inference request through it, then send an outcome report and confirm the loop records it.
Supports weight-training deployments and harness-evolving deployments, so improvement can target the model or its scaffolding.
Your agent sends provider-compatible inference requests through Reef, keeping the integration surface familiar.
Clients send reports about inference outcomes back to Reef, closing the loop that continual improvement needs.
Designed to slot into an agent stack you already run, rather than replacing your agent framework.
Ships recipes and examples for the documented deployment patterns.
Install reef-infra from PyPI, or from source with uv, including an optional slime extra for the training path.
Get the source from the official GitHub repository for reef.
git clone https://github.com/Human-Agent-Society/reef.git
cd reefInstall the published package as the README's installation section shows.
uv pip install reef-infraClone the repository and install it in editable mode (with the training extra if you need it).
git clone https://github.com/Human-Agent-Society/reef.git
cd reef
uv pip install -e .Human-Agent-Society, through the Human-Agent-Society/reef repository, with the site at reefinfra.ai.
Yes — the project is open source under the Apache-2.0 license as recorded for this listing. See the license terms for any conditions on redistribution.
Reef's inference path is provider-compatible, so model calls use your configured provider and keys; Reef itself is free and Apache-2.0 licensed.
Frameworks help you build and run an agent once. Reef is infrastructure for what happens next — collecting inference outcomes and continually improving weights or harness — so it complements a framework rather than competing with it.
Clone the repository (or use the documented installer), follow the install and configure steps on this page, and run the smallest real task first — stand up a minimal deployment, send a provider-compatible inference request through it, then send an outcome report and confirm the loop records it.
A GEP-powered self-evolving engine for AI agents — auditable evolution with Genes, Capsules and Events.
The harness of harnesses: a persistent, self-evolving multi-agent ecosystem built for RSI.
EvoAgentX — a framework for building self-evolving ecosystems of AI agents.
Self-learning browser infrastructure for AI agents — sites remembered, so agents stop rediscovering the web
PraisonAI 🦞 — Hire a 24/7 AI Workforce. Stop writing boilerplate and start shipping autonomous self-improving agents that research, plan, code, and execute t…
Self-evolving memory OS for LLMs and agents: persistent graph memory, hybrid retrieval, skill reuse.
LibreAIカタログに掲載されたAIエージェントプロジェクトの中で最もスターの多いもの — reefが合わなかった場合に検討したい、現実的な候補リストです。
A personal AI that really does things — on any OS, with your data staying yours.
An agentic skills framework & software development methodology that works.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
The agent that grows with you — a Python framework for adaptive AI agents.
DeepSeek's plugin-based agent harness where everything is a plugin.
カタログの中でreefとスター数が近いAIエージェントプロジェクト — 前後の棚もそのまま見て回れます。
Honcho — Plastic Labs' memory library for stateful agents that learn about their users over time.
Put guardrails on LLMs: validate outputs before they reach users.
A batteries-included browser API so agents can automate the web.
OpenTelemetry-native observability for LLM apps.
Flexible and powerful framework for managing multiple AI agents and handling complex conversations
Programmable guardrails to keep LLM systems safe and on rails.