Tencent has open-sourced TencentDB Agent Memory, a fully local memory system designed for AI agents, featuring a 4-tier memory pyramid and hybrid retrieval mechanisms, achieving significant gains in context reduction and accuracy.
Tencent has publicly released TencentDB Agent Memory, a fully local memory system for AI agents under the MIT license. The system implements a sophisticated memory pipeline comprising symbolic short-term memory (offloading tool logs into a Mermaid task canvas) and a 4-tier long-term memory pyramid (L0 Conversation, L1 Atom, L2 Scenario, L3 Persona). It is deployed via an OpenClaw plugin and a Hermes Docker image and defaults to running on local SQLite + sqlite-vec, utilizing hybrid BM25 + vector retrieval with RRF fusion. Performance benchmarks show substantial improvements, including a 61.38% token reduction and a 51.52% relative pass-rate gain on WideSearch with OpenClaw, alongside an increase in PersonaMem accuracy from 48% to 76%.