OpenCLAWN Agent Framework

Skill by ara.so — Hermes Skills collection.

OpenCLAWN is a lightweight, self-improving multi-agent framework built around 4 core innovations: routing audit + self-calibration, skill decay, confidence-gated crystallization, and role output contracts. It features hybrid local (Ollama) + cloud (Gemini/Claude) LLM routing, 26 sandboxed tools, and a compounding skill library that tidies and improves itself as it's used.

Key capabilities:

• Multi-agent conversations (pipeline, debate, orchestrator strategies)
• Self-calibrating smart router with multilingual complexity detection
• Skill compounding: promote, refine, merge, decay — all versioned & revertible
• Autopilots with approval-gated proposals (no silent execution)
• Activity timeline tracking every agent action
• Skill pack export/import with SSRF + injection guards

Installation

Prerequisites