Memory

关于

Memory provides a local-first MCP server for agent memory, designed to separate fast capture, runtime retrieval, and maintenance curation processes. It uses SQLite with `sqlite-vec` for efficient embedded storage and search, leveraging OpenAI embeddings for semantic understanding. Agents can log candidate memory during their normal operations, which is then processed by a maintenance routine. This routine converts relevant log items into persistent, searchable facts, ensuring a structured and curated memory for AI agents.

主要功能

• Configurable OpenAI embeddings for semantic memory
• Separation of fast capture, runtime retrieval, and maintenance curation
• 0 GitHub stars
• Queue-based maintenance flow for memory logging and acknowledgement
• Embedded storage and search with SQLite + sqlite-vec
• CRUD tools for comprehensive memory management

使用案例

• Logging candidate memory from AI agents during their operational tasks
• Enabling runtime retrieval of relevant facts for agent decision-making
• Curating and compacting agent memory through a structured maintenance process