MemoriLabs/Memori v3.3.3: Agent-Native Memory Infrastructure Launched May 7 at 14.3K Stars, Beats Zep and Mem0 on LoCoMo Benchmark

GitHub·high signal

Memori is an LLM-agnostic memory layer that turns agent execution traces — tool results, workflow steps, decision logic — into structured persistent state. On the LoCoMo benchmark it achieved 81.95% accuracy using only 1,294 tokens per query (4.97% of full-context), outperforming Zep, LangMem, and Mem0 while reducing prompt size by 67% vs Zep. Officially launched May 7 with Python and TypeScript SDKs, the repo has surged to 14.3K stars with 549 commits.