← Back to Discover

TencentDB Agent Memory

Open-source memory hub that lets teams of AI agents keep, share and reuse what they learn from conversations, documents and code across sessions and frameworks.

Vector databases & RAGNotes & knowledge
Visit website ↗

What it does

TencentDB Agent Memory is a self-hosted memory layer for teams that run several AI agents. It stores what agents learn so the next session, or the next agent, does not start from nothing, and it keeps those memories separate from any one agent framework. Four kinds of stored memory are supported: chat memory distilled in layers from past conversations, reusable skills extracted from finished work, wiki pages built from documents, and a code index covering symbols, call relationships and the impact of a change. Existing repositories, documents and past conversation sessions can be imported so a new agent has a head start. A web panel handles teams, review and sharing. New chat memory and skills are private by default, and visibility can be widened to a team or restricted with user, role or agent access rules. Memory assets can be bound to individual agents so each one receives only what it needs. Agents connect through a proxy by pointing their base URL at it, with no plugin, hook or MCP server required. The site lists Claude Code, Codex, CodeBuddy, DeepSeek Harness, WorkBuddy, Hermes and OpenClaw as supported clients. It notes that Wiki and CodeGraph are built asynchronously and that CodeGraph currently prioritises public HTTPS repositories.

Capabilities

  • Chat memory distilled in four layers, from raw conversation to long-term persona
  • Skill library extracted from conversations and tool calls, with review before sharing
  • Wiki pages with a link graph and a code index of symbols and call relationships
  • Private, team and restricted visibility with user, role and agent access rules
  • Proxy-based setup for Claude Code, Codex, CodeBuddy and other agents
  • Import of existing repositories, documents and past sessions for a faster start

When to use it

  • A team runs several agents and wants them to share experience
  • Agents keep re-learning project context at the start of every session
  • You want existing docs, code and chat history turned into material agents can look up
  • You need control over who, and which agent, can see each stored memory