A local-first, SQLite-backed memory toolkit for AI agents that uses FTS5 search and a five-tier schema instead of vector databases or embeddings.
Adapted from @DanKornasYour agent memory doesn’t need a vector database. Sibyl Memory is a file-based long-term memory toolkit for builders adding durable memory to AI agents. It helps you store and retrieve agent context locally by using SQLite, FTS5 search, and a five-tier memory schema instead of embeddings or an external retrieval service. Key features: • Local-first storage – memory lives in a single SQLite file on the agent’s machine • FTS5 search – search stored memory without a vector database or embedding model • Five-tier schema – separates live state, entities, journal entries, reference material, and archived records • Multiple integration paths – includes an SDK, CLI, MCP server, Hermes integration, and LangGraph BaseStore adapter • Multi-tenant design – the client SDK supports tenant-aware memory storage It’s open-source (MIT license). Link in the reply 👇 🔗 GitHub: https://github.com/Sibyl-Labs/Sibyl-Memory --- ✉️ If you’re into AI, ML, agents, and building real systems, join my newsletter (it’s free): https://dankornas.substack.com
