Provides a high-performance, persistent, semantic, and graph-based memory server for AI agents.
Sponsored
Memory is a pure Rust, 'all-in-one' Model Context Protocol (MCP) server designed to give AI agents long-term memory capabilities. Unlike complex setups, it's a single, self-contained executable that embeds SurrealDB, runs embedding models locally via ONNX, and requires zero external dependencies or API keys. It integrates vector search for semantic similarity, a knowledge graph for entity relationships, and code indexing for codebase understanding, using hybrid retrieval for optimal results.
主な機能
01Single, self-contained executable with embedded database (SurrealDB)
020 GitHub stars
03Local embedding models (FastEmbed) run on CPU, no API keys required
04Combines vector search, knowledge graph, and code intelligence
05Implements agent protocols for context window compaction and session restarts
06Exposes 26 tools for memory management, search, graph interaction, and codebase intelligence
ユースケース
01Integrating long-term memory with AI IDEs like Claude Desktop, Cursor, and OpenCode
02Enabling AI agents to maintain context across sessions and context window resets
03Providing AI agents with semantic understanding of codebases and entity relationships
