Sign InOpen Brain
GitHubGitHub RepoNeeds Review

TencentCloud/TencentDB-Agent-Memory

An open-source memory hub turns agent conversations, workflows, docs, and code into governed assets that can be reused across sessions and roles, reducing repeated project setup.

GitHub
Open Source Open MarkdownOpen JSON
Source Summary

TencentDB Agent Memory packages prior work into **four memory assets**: Chat Memory, Skills, Wiki, and CodeGraph. The hub adds ownership, versions, agent bindings, and private, team, or ACL-based access.

Practical Implication

Use it to give coding and review agents different context loadouts instead of one global prompt. Memory is distilled through **L0–L3**, while detailed recall combines **BM25, vector retrieval, and RRF** under context-size and timeout limits.

Agent-Ready Context
TencentDB Agent Memory packages prior work into **four memory assets**: Chat Memory, Skills, Wiki, and CodeGraph. The hub adds ownership, versions, agent bindings, and private, team, or ACL-based access.

Use it to give coding and review agents different context loadouts instead of one global prompt. Memory is distilled through **L0–L3**, while detailed recall combines **BM25, vector retrieval, and RRF** under context-size and timeout limits.

The Team Memory release is still a beta. Wiki and CodeGraph indexing is asynchronous, private-repository support and automatic routing remain incomplete, and current compatibility is limited to **OpenClaw, Hermes, and SDK integrations**.
Connected Context · Feed7 Judgment

This turns the proposed shared context layer into a concrete, permissioned asset hub with agent-specific loadouts, versioning, distilled levels, and bounded hybrid retrieval. It broadens memory beyond chat history to skills, documentation, and code structure, while narrowing near-term adoption through beta status, asynchronous indexing, incomplete routing and private-repository support, and limited integrations.

WTF Is the Context Layer? The Missing Infrastructure for Production Agents — Prukalpa SankarTencentDB implements much of Atlan’s proposed shared, versioned context layer by adding ownership, access controls, agent bindings, retrieval, and multiple reusable context asset types.UniMem: Complementary Episodic-to-Parametric Memory for Boundary-Agnostic Task StreamsTencentDB keeps knowledge in inspectable retrieval assets, whereas UniMem also consolidates recurring experience into model parameters; together they expose a design choice between governed external memory and less directly inspectable learned memory.virgiliojr94/book-to-skillbook-to-skill produces portable skills from segmented documents; TencentDB provides a governed hub where skills can coexist with wiki and code-graph assets and be assigned selectively to agents.EveryInc/compound-engineering-pluginCompound Engineering persists workflow artifacts as future context, while TencentDB supplies ownership, versions, permissions, and retrieval machinery that could govern such durable team memory across agents.
Context Map
agentcoding#agent-memory#skills#context-engineering
Uncertainty
The Team Memory release is still a beta. Wiki and CodeGraph indexing is asynchronous, private-repository support and automatic routing remain incomplete, and current compatibility is limited to **OpenClaw, Hermes, and SDK integrations**.