TencentDB Agent Memory
ActiveDescription
TencentDB Agent Memory delivers fully local long-term memory for AI agents via a 4-tier progressive pipeline, with zero external API dependencies, vector search, and embedding storage.
Key Features
- Symbolic short-term memory using Mermaid canvas to compress verbose logs into compact tokens
- Layered long-term memory with 4-tier semantic pyramid (L0 Conversation → L3 Persona)
- Up to 61.38% token reduction and 51.52% pass rate improvement with OpenClaw
- Full traceability with deterministic drill-down from high-level abstractions to raw evidence
- Heterogeneous storage: database for facts, Markdown for personas and scenes
- Zero external API dependencies, fully local deployment with vector search and embeddings
Use Cases
💡 Long-horizon AI agent sessions with context accumulation
💡 Enterprise workflow automation requiring persistent task context
💡 Reducing token costs in production agent deployments
💡 Building personalized AI assistants that remember user preferences
💡 Multi-turn conversation systems with structured memory recall
Strengths & Limitations
✅ Strengths
- • Actively maintained, recent updates
- • High community interest (25.7k stars)
⚠️ Limitations
- • No clear open-source license
Categories
Quick Start
Install: npm install @tencentdb-agent-memory/memory-tencentdb. Requires Node >= 22.16. Integrate with OpenClaw (>= 2026.3.13) or Hermes Gateway for automatic memory management. The plugin handles symbolic offloading and layered memory retrieval automatically.