TencentDB-Agent-Memory: TencentDB Agent Memory

TencentDB Agent Memory

What is TencentDB Agent Memory?

TencentDB Agent Memory is a system that enables AI agents to retain and reuse experience (chat memory, skills, wiki, codegraph) across sessions and team members to reduce repetitive work.

Purpose

Eliminates the need for agents to repeatedly learn project context by enabling memory asset sharing and reuse across sessions and team members.

How It Works

Extracts Chat Memory, Skills, Wiki, and CodeGraph from conversations, documents, and code; manages them in a Memory Hub with access controls (private/team/ACL); agents retrieve relevant assets on demand via layered memory (L0-L3) and tool calls to avoid context window overload.

Key Features

  • Automatic asset extraction: Chat Memory/Skills from convos, Wiki from docs, CodeGraph from code
  • Framework-portable memory assets usable across agent frameworks
  • Cold-start: import existing codebases, documents, conversation sessions
  • Team memory panel: create teams, manage assets, control access (private/team/ACL), bind assets to agents
  • Layered memory (L0-L3) for efficient recall: L0 (raw), L1 (facts), L2 (scenarios), L3 (persona)
  • Versioned Skill library with triggers, resource steps, validation
  • Wiki for structured documentation and link graphs
  • CodeGraph for code symbol indexing, call relationships, impact analysis
  • Human-controlled asset ownership, versioning, status, visibility

Use Cases

  • Building agent teams for solo entrepreneurs (e.g., Scout, Builder, Reviewer roles)
  • Avoiding retraining via "save file" of existing experience (codebase, docs, convos)
  • Sharing specific skills (e.g., release checklist, troubleshooting) across agents
  • Using Wiki to access structured docs without reading every file
  • Using CodeGraph to analyze code impact before changes

Limitations

  • Wiki/CodeGraph require async processing time to reach ready state
  • CodeGraph currently prioritizes public HTTPS repos; private repo/SSH support still being refined
  • Memory Hub supports manual asset binding; fully automated routing under iteration
  • Currently supports OpenClaw, Hermes, Claude Code, CodeBuddy, SDK; broader framework integration on roadmap

Tech Stack

  • Node.js (>=22.16)
  • CodeGraph (dependency for CodeGraph asset module)
  • Hermes Agent (dependency for Skill asset management)

Sources