Provides AI agents with a self-hosted, persistent, and semantically searchable memory server.
Sponsored
Eidetic is designed as a semantic memory server for AI agents, enabling them to retain and retrieve information without relying on external cloud services. It's built with a focus on self-hosting, ensuring your data remains on your hardware, and is CPU-friendly, running comfortably on minimal resources as it doesn't require a GPU for embeddings. Leveraging PostgreSQL with the pgvector extension, it offers powerful semantic search capabilities through cosine similarity, going beyond simple keyword matching. Every memory is also backed up to disk as a human-readable Markdown file, providing a robust flat-file backup system. It communicates via the standard MCP protocol, making it compatible with various AI agent frameworks out of the box.
主な機能
01Standard MCP protocol (JSON-RPC) for agent compatibility
020 GitHub stars
03CPU-friendly, no GPU required for embeddings
04Pgvector-powered semantic search for relevant memory retrieval
05Self-hosted memory server for AI agents
06Flat-file backup of all memories as human-readable Markdown
ユースケース
01Integrating semantic memory capabilities into MCP-compatible AI agent frameworks
02Enabling AI agents to recall past interactions and learned information efficiently
03Providing AI agents with persistent, searchable memory without cloud dependencies
