OpenSpace is an open-source self-evolving AI agent skill engine developed by HKUDS (the Data Intelligence Lab at the University of Hong Kong), the team behind LightRAG. Its tagline is 'One Command to Evolve All Your AI Agents.' The project addresses three agent pain points: token waste from repeated reasoning, repeated errors that agents cannot share, and skill decay as tools change. OpenSpace treats skills as living entities that evolve through three modes — FIX, DERIVED, and CAPTURED — triggered by post-execution analysis, tool degradation detection, and metric monitoring. Its collective intelligence layer connects agents via a cloud skill community, so one agent's improvement benefits all. On the GDPVal benchmark (50 real workplace tasks, same backbone model Qwen 3.5-Plus), OpenSpace agents cut token consumption by 46% and improved revenue 4.2x versus baseline. Safety features include confirmation gates, loop guards, and diff-driven minimal patches stored in a version DAG with full lineage tracking. The article also compares OpenSpace with MetaClaw, showing they are complementary: OpenSpace is a plug-in skill-evolution layer, while MetaClaw performs weight-level continual learning.
OpenSpace Deep Dive: A Self-Evolving Skill Engine That Makes All Your AI Agents Smarter
Summary
OpenSpace is an open-source self-evolving AI agent skill engine developed by HKUDS (the Data Intelligence Lab at the University of Hong Kong), the team behind LightRAG. Its tagline is 'One Command to Evolve All Your AI Agents.' The project addresses three agent pain points: token waste from repeated reasoning, repeated errors that agents cannot share, and skill decay as tools change. OpenSpace treats skills as living entities that evolve through three modes — FIX, DERIVED, and CAPTURED — triggered by post-execution analysis, tool degradation detection, and metric monitoring. Its collective intelligence layer connects agents via a cloud skill community, so one agent's improvement benefits all. On the GDPVal benchmark (50 real workplace tasks, same backbone model Qwen 3.5-Plus), OpenSpace agents cut token consumption by 46% and improved revenue 4.2x versus baseline. Safety features include confirmation gates, loop guards, and diff-driven minimal patches stored in a version DAG with full lineage tracking. The article also compares OpenSpace with MetaClaw, showing they are complementary: OpenSpace is a plug-in skill-evolution layer, while MetaClaw performs weight-level continual learning.
This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/177169501