Multi-Agent Harness for Production AI
The Harness of Harnesses • built for RSI: a trusted, persistent, self-evolving multi-agent ecosystem for all-domain collaboration.
构建受监督的、自我进化的 Agent 组织的基础设施 | Infrastructure for supervised, self-improving agent organization. 飞书/Telegram 手机端运行 Claude Code 或 Kimi Code（双引擎，两家原生订阅直接用），共享记忆、Agent 工厂、定时任务、通信总线。
A curated research map of Recursive Self-Improvement (RSI): models, agents, harnesses, embodied systems, automated AI R&D, benchmarks, and safety.
The official repository of "Position: Agentic Evolution is the Path to Evolving LLMs".
Turn the agents you already use into a squad. Codex, Claude Code, Cursor, Hermes Agent and OpenClaw use one encrypted memory, pass work to each other, review each other's work, and dream every night to wake up a little better. Open source. Runs on your computer.
Sno Station Skills: 31 tested skills that turn Claude Code and Codex into one working team. Cross-vendor peer review, automatic handoff when an agent hits its quota, a one-page morning brief, work orders with proof, and a nightly loop that learns from your own sessions. Open source, installed with one sno setup.
My AI Stand. Realtime by day, rewriting itself by night. Summon my AI superpower.
The agent that grows with you
Kapso by Leeroo: Long-running agents that optimize AI and Data systems, and learn from every experience. #1 open-source on MLE-Bench; ALE-Bench; RelBench
ClawCode is claude code inspired implementation in python and Rust focused on agents and experience-based evolution, and is an Open-source coding-agent CLI for Anthropic, OpenAI, Gemini, DeepSeek, GLM, Kimit, Ollama, Codex, GitHub Models, and 200+ models via OpenAI-compatible APIs.
SiriuS: Self-improving Multi-agent Systems via Bootstrapped Reasoning
🔨 Kyoko is the all-in-one, fully local tool for debugging and improving your AI agents.
Self-correcting learning engine for Claude Code — persistent identity, behavioral pattern tracking, and cross-session memory
Self-Evolved Diverse Data Sampling for Efficient Instruction Tuning
A task conductor for Claude Code that gets better the more you use it. Multi-agent orchestration with an evidence contract that BLOCKS instead of asking, gates written as hooks rather than prompts, and a retrospective that learns your repo's rules and tunes its own playbook.
A tiny model that teaches itself to code better. On your laptop. No cloud. No teacher model. No human feedback.
The coding agent you don’t have to watch. Start a task and walk away. AFK builds the feature, verifies its own work, and texts you when it’s done. You come back to a receipt of every step, every decision, and what’s waiting on you. If it’s unsure, it stops and asks instead of guessing.
Detail-Oriented CLIP for Fine-Grained Tasks (ICLR SSI-FM 2025)
Self-improving code for any git repo: give it a measurable goal, a coding agent (Claude Code or Aider) proposes changes, the kernel scores them, keeps winners and rolls back failures. Open-source AlphaEvolve / FunSearch-style evolution loop, sandboxed, with an auditable ledger.
To associate your repository with the self-improving topic, visit your repo's landing page and select "manage topics."