Tencent Open-Sources TencentDB Agent Memory: A 4-Tier Local Memory Pipeline for AI Agents
61% token reduction. Tencent just open-sourced a local memory system that lets AI agents think without burning context.
Why it matters
Tencent's TencentDB Agent Memory addresses a critical pain point for agent builders: context window bloat from tool logs and conversation history. By separating symbolic short-term memory from a tiered long-term pyramid, it cuts token costs while improving reasoning accuracy—a practical infrastructure move that could shift how teams architect production agents.
The key facts
8 to knowMIT-licensed open-source release
4-tier memory architecture: L0 Conversation → L1 Atom → L2 Scenario → L3 Persona
61.38% token reduction on benchmarks
51.52% relative pass-rate gain on WideSearch with OpenClaw
PersonaMem accuracy improvement: 48% → 76%
Hybrid BM25 + vector retrieval with RRF fusion
Ships with OpenClaw plugin and Hermes Docker image
Local SQLite + sqlite-vec backend by default
Go to the source
MarkTechPostmarktechpost.com
Publisher excerpt: Tencent has open-sourced TencentDB Agent Memory, a fully local memory system for AI agents released under the MIT license. The project pairs symbolic short-term memory, which offloads verbose tool logs into a compact Mermaid task canvas, with a 4-tier long-term memory pyramid (L0 Conversation → L1…