AI/TLDR

TencentDB Agent Memory

A team-level memory hub that turns a team's conversations, docs and code into four governed assets every agent can reuse

Agent MemoryOpen source
Latest
v2.0.0
Updated
3 Aug 2026
Language
TypeScript
License
MIT
Coverage
1 story

What's new

v2.0.03 Aug 2026

Tencent Cloud open-sourced v2.0, which turns team conversations, docs and code into four reusable assets shared across agents through Anthropic- and OpenAI-compatible endpoints.

Latest news

Overview

TencentDB Agent Memory is a memory hub built for teams rather than for a single developer's session. It takes the raw output of everyday work — conversations, documents and code — and distils it into four reusable assets: Chat Memory, a Skill library, an LLM-Wiki and a Code-Graph. Those assets are owned, versioned and permissioned, so an agent that joins a project inherits what the team already knows instead of starting from an empty context.

Chat Memory distils conversations through multiple layers (L0–L3) to capture preferences, decisions and context. The Skill library extracts versioned workflows out of completed tasks. The Wiki holds structured documentation with a link graph, and CodeGraph indexes symbols, files, call relationships and impact analysis. A team control panel governs all four with ownership, versioning and private/team/restricted roles.

One proxy serves many harnesses: the README lists DeepSeek Harness, Claude Code, Codex, CodeBuddy, WorkBuddy, Hermes and OpenClaw among the supported clients, so several agents can read and write the same memory. It runs as a self-hosted Docker deployment on Node 22.16 or later, and is MIT licensed.

What it does

  • Four memory assets — Chat Memory, Skills, LLM-Wiki and Code-Graph — built out of a team's own work
  • Multi-layer conversation distillation (L0–L3) that keeps preferences, decisions and context rather than raw transcripts
  • CodeGraph indexes symbols, files and call relationships for impact analysis across a repository
  • Team control panel with ownership, versioning and private / team / restricted access roles
  • One proxy equips many harnesses, including Claude Code, Codex, DeepSeek Harness, CodeBuddy and OpenClaw
  • Self-hosted Docker deployment with a web interface on localhost

Getting started

The project ships a Docker deployment under deploy/global-images. Configure the environment file, then bring the whole stack up with one script.

Clone the repository

Everything needed for a self-hosted deployment is in the repo.

bashbash
git clone https://github.com/TencentCloud/TencentDB-Agent-Memory.git
cd TencentDB-Agent-Memory/deploy/global-images

Configure the environment

Copy the example environment file and fill in your model and database settings before starting.

bashbash
cp .env.example .env
$EDITOR .env

Start the stack

The start script brings up the services; the web interface is then served locally.

bashbash
./start-all.sh
# open http://localhost:8125

Point your agents at it

A single proxy serves the supported harnesses — Claude Code, Codex, DeepSeek Harness, CodeBuddy, WorkBuddy, Hermes and OpenClaw — so they share one memory instead of keeping private ones.

texttext
# configure your harness to use the TencentDB Agent Memory proxy endpoint

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

When to use it

  • Give a whole engineering team one shared agent memory instead of a private context per developer
  • Turn finished tasks into a versioned skill library the next agent can execute
  • Let an agent reason about a large codebase through an indexed call graph and impact analysis
  • Govern what agents may read with ownership and private/team/restricted roles

How TencentDB Agent Memory compares

TencentDB Agent Memory alongside other open-source agent memory tools AI/TLDR tracks, ranked by GitHub stars.

ToolStarsWhat it does
Claude-Mem★ 93.6kPersistent memory layer for coding agents: it captures what an agent does during a session, compresses it into semantic summaries, and injects the relevant parts back into later sessions.
Mem0★ 65.1kA memory layer that you add to existing LLM agents to extract, store, and recall user facts and preferences across sessions using vector, graph, and key-value backends.
MemPalace★ 59kLocal-first agent memory that stores conversations verbatim instead of summarising them, with a structured palace index, pluggable vector backends and an MCP server.
OpenViking★ 36.5kA context database that stores an agent's memories, resources, and skills as one browsable viking:// filesystem with three-tier (abstract/overview/details) on-demand loading.
Graphiti★ 30.8kA library that builds a temporal knowledge graph from an agent's conversations and data so facts can be tracked and queried as they change over time.
Cognee★ 30.6kA graph-native memory engine that turns raw documents and conversations into a queryable knowledge graph for agents that need to build lasting knowledge.
Supermemory★ 29.6kA memory and context engine that ingests information across tools and sessions and can run fully locally, acting as a second brain for AI applications.
TencentDB Agent Memory★ 26.3kA team-level memory hub that turns a team's conversations, docs and code into four governed assets every agent can reuse