Metronix Memory is a self-hosted memory infrastructure and MCP server designed to provide AI agents with durable, long-term recall and contextual understanding. By combining dense vector search via Qdrant, SPLADE sparse retrieval, and Neo4j graph context, it enables agents to persist facts, capture user preferences, and navigate complex multi-hop knowledge. Fully containerized with Docker Compose, it ensures complete operational control over your data while natively integrating with tools like Claude Desktop, Cursor, and VS Code through MCP or standard REST APIs.
主要功能
01107 GitHub stars
02Durable agent memory with workspace and per-agent scoping for facts, preferences, and pinned context
03Hybrid knowledge retrieval combining dense vectors, SPLADE sparse search, and Neo4j graph traversal with source citations
04Self-hosted Docker Compose architecture backed by PostgreSQL, Qdrant, Neo4j, and Redis
05Native MCP memory server alongside REST and OpenAI-compatible API interfaces
06Automated knowledge freshness pipeline to detect stale, conflicting, or updated data
使用案例
01Cross-session memory persistence for developer tools like Cursor, Claude Code, and VS Code
02Enterprise knowledge retrieval and multi-hop reasoning over private documentation and SaaS data
03Multi-agent coordination with role-scoped facts, shared workspaces, and isolated agent contexts