Tencent Releases TencentDB-Agent-Memory, a Shared Memory Hub for AI Agents
Tencent Cloud has open-sourced TencentDB-Agent-Memory, a team-level memory hub designed to manage and share memory assets across different AI agent frameworks. The repository has quickly gained significant popularity, accumulating over 10,000 stars within a single month. Memory management and persistence remain critical bottlenecks for AI agents, and this tool provides a structured way to share context, skills, and code graphs across multiple agents. By enabling team-level memory sharing, it helps developers build more collaborative and consistent multi-agent systems. Written in TypeScript, the project organizes conversations, documents, and code into four reusable assets: Chat Memory, Skill, LLM-Wiki, and Code-Graph. These assets can be governed and equipped across various agent frameworks to maintain state and knowledge.
## BACKGROUND
AI agents typically rely on short-term memory (like prompt context) and long-term memory (like vector databases) to perform tasks. However, sharing this memory across different agents or team members in a standardized format is challenging. Tools like Mem0 and LangGraph also address agent memory, but Tencent's solution focuses specifically on a unified team-level hub for multiple asset types.