Facebook (Md Ismail Sojal)
ai-engineeringai-tools
Original source

TencentDB Agent Memory — Fully Local Long-Term Memory for AI Agents

Summary

  • Open-source local-first memory system by TencentCloud that gives AI agents human-like long-term recall
  • No cloud dependency, no API bills — runs fully local
  • Turns conversations, docs, and code into 4 reusable memory types: Chat Memory, Skill, LLM-Wiki, Code-Graph
  • Team-level memory hub — governed, shared, and equipped across agents and frameworks
  • Reported benchmarks:
    • 61.38% fewer tokens per session
    • 76% persona accuracy (up from 48%)
    • 51.52% higher task success rate
  • Works with any agent framework
  • Written in TypeScript
  • Topics: agent, embedding, LLM, local-first, long-term-memory, vector-search