Tencentdb Agent Memory
- 6 repo stars
- Updated July 26, 2026
- baodq97/tencentdb-agent-memory
tencentdb-agent-memory is a Claude Code plugin that stores four-layer long-term memory locally with hybrid FTS5 and vector recall so agents can remember context across sessions without external APIs.
About
tencentdb-agent-memory is a Claude Code plugin inspired by TencentDB Agent Memory that implements four-layer long-term memory entirely on your machine. Hooks inject hybrid FTS5 and vector recall on each prompt, auto-capture turns, and consolidate scenes and persona without external APIs. Use it when you want durable, privacy-preserving Claude Code memory with a tmem CLI and optional GitHub contributor intelligence via /contrib.
- Four-layer memory: L0 conversation, L1 atoms, L2 scenes, L3 persona
- Hybrid FTS5 plus EmbeddingGemma vector search merged with RRF
- Fully local with no external gateway or paid API required
- tmem CLI for search, recall, scenes, sync, and daemon management
- /contrib profiles GitHub engineers into evidence-linked contributor personas
Tencentdb Agent Memory by the numbers
- Data as of Jul 27, 2026 (Skillselion catalog sync)
/plugin marketplace add baodq97/tencentdb-agent-memoryAdd your badge
Show developers this marketplace is listed on Skillselion. Paste this into your README.
| repo stars | ★ 6 |
|---|---|
| Last updated | July 26, 2026 |
| Repository | baodq97/tencentdb-agent-memory ↗ |
How can Claude Code retain and retrieve relevant memories across sessions without cloud memory services or paid embedding APIs?
Give Claude Code four-layer local long-term memory with hybrid FTS5 and vector recall plus optional GitHub contributor profiling.
Who is it for?
Developers running Claude Code who want local, hybrid-recall long-term memory with CLI inspection and optional GitHub contributor profiling.
Skip if: Teams needing hosted multi-user memory sync or simple one-shot context without setup of local embedding models.
What you get
Creates local memory stores under ~/.memory-tencentdb, injects recalled persona and atoms into prompts, and exposes tmem commands for search, scenes, and consolidation.
Plugins in this marketplace
1 plugin - install individually after you add the marketplace.
Recommended Marketplaces
FAQ
Does it require external APIs?
No. All extraction and consolidation is done by the Claude agent with local FTS5, sqlite-vec, and EmbeddingGemma-300m embeddings.
What happens automatically each turn?
UserPromptSubmit runs hybrid recall and injects persona, relevant atoms, and an L2 scene-navigation index; Stop and SessionEnd capture and queue consolidation.
What is /contrib for?
It profiles public GitHub engineers into evidence-linked contributor personas and playbooks stored separately from self-memory.