Show HN: Rekal – Long-term memory for LLMs in a single SQLite file

Category: devtools

Tags: llm-memory, sqlite, mcp-server, local-ai, vector-search

Score: 7.3/10 (Innovation: 7, Technical: 7, Documentation: 8, Utility: 7)

Rekal is a local, SQLite-based long-term memory system for LLMs that implements hybrid search (keywords + vectors + recency) via the Model Context Protocol. It's interesting because it provides persistent memory for AI assistants without cloud dependencies, using a single file with sophisticated conflict detection and conversation-aware storage.

Target audience: backend devs, ai engineers, developers using ai coding assistants

Repository: https://github.com/janbjorge/rekal · Python · MIT

View on Hacker News