TencentDB Agent Memory

TencentDB Agent Memory is a team-level memory hub for AI agents, built by Tencent Cloud. It turns the raw material agents produce every day (conversations, documents and code) into governed, reusable memory assets that can be shared across agents, frameworks and team members instead of being lost when a session ends.

Overview

The project has gathered more than 29,700 stars on GitHub and is under intense active development, with updates in October 2026. It organizes agent knowledge into four asset types: Chat Memory for conversation history, Skills for reusable procedures an agent has learned, an LLM-Wiki for structured knowledge, and a Code-Graph for understanding codebases. Local-first storage with vector search keeps retrieval fast, while governance features control what is shared and with whom. It ships as a plugin for agent frameworks such as OpenClaw, so an existing agent can gain durable, team-shared memory without being rebuilt, and embeddings power semantic recall across everything stored.

Key Features

• Four reusable memory assets: chat memory, skills, LLM wiki and code graph
• Team-level sharing and governance instead of per-session, per-agent silos
• Local-first architecture with vector search for fast semantic recall
• Plugin integration for OpenClaw and other agent frameworks
• Turns conversations, documents and code into searchable long-term memory
• From Tencent Cloud, designed for production multi-agent deployments
• Continuously updated open-source release with enterprise backing

💳 Pricing

The agent memory components are free to use as open-source plugins in local and self-hosted setups. Managed TencentDB cloud services and enterprise deployments from Tencent Cloud are paid.

Who Is It For

TencentDB Agent Memory is for teams running multiple AI agents that repeat the same mistakes and rediscover the same knowledge, and for platform engineers who want shared, governed, long-term memory as infrastructure rather than a per-bot hack.