Agent Memory MCP

Semantic vector memory for AI agents. Local LanceDB storage with hybrid BM25 + vector search using all-MiniLM-L6-v2.

active
mcpmemoryvector-searchlancedb

Agent Memory MCP provides persistent semantic memory for AI agents. Store experiences, patterns, and learnings; retrieve them later by meaning rather than exact keyword match.

The server runs entirely locally — no cloud APIs. Embeddings are generated with all-MiniLM-L6-v2 (ONNX runtime), stored in LanceDB, and searchable via hybrid BM25 + vector scoring. Categories and tags provide additional structure. Hardcopy JSON backup ensures nothing is lost.

Install via npm:

npm install -g @adamrdrew/agent-memory-mcp