|

Tencent Cloud Open-Sources TencentDB Agent Memory v2.0: A Team-Level Memory Hub for AI Coding Agents

ByRicardo

Tencent Cloud has open-sourced TencentDB Agent Memory v2.0, a team-level reminiscence hub for AI brokers. The concept is tremendous easy: if venture context was already defined as soon as, a brand new session mustn’t want it repeated. The system turns conversations, paperwork and code into 4 reusable reminiscence property — Chat Memory, Skill, LLM-Wiki and Code-Graph — which might be versioned, permissioned and outfitted to particular brokers. Single-agent reminiscence will not be new. What is new right here is the governance layer, which lets a teammate’s agent learn what your agent discovered, with out leaking something you marked non-public. The steady 2.0.0 launch is printed on August 3, 2026.

Is it deployable?

Yes, it’s deployable. TencentDB Agent Memory is MIT-licensed and self-hosted. Three Docker photos printed to Docker Hub begin with one command, and multi-arch builds cowl linux/amd64 and linux/arm64.

  • Which firms: Solo builders and small engineering groups get probably the most worth now. The venture explicitly targets the one-person firm. Mid-size orgs with a platform or DevEx operate can run it as shared infrastructure. Large regulated enterprises ought to pilot somewhat than standardize, as a result of private-repo CodeGraph and automatic reminiscence routing are nonetheless being refined.
  • Industries: Software and developer tooling, SaaS, fintech, consulting and businesses, plus any regulated group that wants reminiscence to remain inside its personal community.
  • Applications: Onboarding a brand new agent to an present codebase, influence evaluation earlier than refactoring, launch checklists, incident runbooks, code evaluate requirements, and turning product docs into agent-readable pages.

Four reminiscence property

The system converts work into 4 asset sorts:

  • Chat Memory retains preferences, information, selections and interplay historical past.
  • Skill distills reusable procedures from accomplished duties, carrying variations, useful resource recordsdata, set off boundaries, execution steps and validation guidelines.
  • Wiki turns paperwork into structured pages with a hyperlink graph, an method knowledgeable by Andrej Karpathy’s LLM-maintained knowledge base idea.
  • CodeGraph indexes symbols, recordsdata, name relationships and influence paths.

All 4 register uniformly as Memory Assets, so possession, model, standing and visibility behave identically throughout them.

Layered distillation, budgeted retrieval

Chat Memory will not be flat. Conversations save as L0, then an async pipeline refines them into L1 Atom, L2 Scenario and L3 Core/Persona. Retrieval is layered too. L2 and L3 present a quick context bootstrap. When particular information are wanted, BM25 plus vector retrieval plus RRF falls again to L1 and L0. Results are capped by merchandise depend, character funds and timeout, so reminiscence doesn’t crowd out the context window.

Governance is the precise differentiator

Standard RAG solutions what may be discovered. The Hub additionally solutions who might use it, which model is legitimate, and which agent receives it. Visibility runs non-public, group and restricted, with agent for focused equipping. Per the README, non-public is owner-only, not readable even by group admins. New Chat Memory and Skills default to non-public, making sharing an express motion.

Memory Hub makes use of fastened binding plus ACL: slender by group, consumer, agent and visibility first, then retrieve.

Wiring brokers in

Memory Proxy speaks each Anthropic and OpenAI protocols, exposing /claude-code/<spaceId>/v1/messages and /v1/chat/completions. On the primary flip, sessionInit makes use of Claude Code’s native AskUserQuestion device to select group, agent and activity. Every subsequent flip injects that agent’s L2/L3 reminiscence, matched abilities and information into the system immediate earlier than forwarding upstream.

Default ports are 8420 for Memory Core, 8125 for the panel, 8424 for the information service and 8096 for the proxy, per INSTALL.md. Official SDKs ship for TypeScript and Python. Supported integrations are OpenClaw, Hermes, Claude Code, CodeBuddy and direct SDK use.

The 2.0.0 launch provides Skill compelled archiving, scheduled CodeGraph repository sync, system-admin asset administration, English/Chinese panel switching, and a Cost Guard that assigns cheaper fashions to particular brokers.

Reported outcomes

It stories PersonaMem accuracy rising from 48% to 76%, a 59% relative enchancment. That determine is self-reported, and no impartial replica has been printed.

Key Takeaways

  • MIT-licensed, self-hosted, one-command Docker deploy — no vendor API dependency.
  • Four asset sorts exchange chat logs: Chat Memory, Skill, Wiki, CodeGraph.
  • L0→L3 distillation with BM25 + vector + RRF retrieval below strict funds caps.
  • ACL-governed sharing is the actual differentiator over customary RAG.
  • PersonaMem 48%→76% is self-reported; private-repo CodeGraph is unfinished.

Check out the GitHub repository, CHANGELOG, INSTALL.md, and Knowledge OpenAPI. Also, be happy to observe us on Twitter and don’t neglect to hitch our 150k+ML SubReddit and Subscribe to our Newsletter. Wait! are you on telegram? now you can join us on telegram as well.

Need to companion with us for selling your GitHub Repo OR Hugging Face Page OR Product Release OR Webinar and so on.? Connect with us

The publish Tencent Cloud Open-Sources TencentDB Agent Memory v2.0: A Team-Level Memory Hub for AI Coding Agents appeared first on MarkTechPost.

Similar Posts