Kioku Lite – Knowledge Graph memory for AI agents in a single SQLite file

Summary: Kioku Lite is an offline memory engine for AI agents that combines keyword search, vector similarity, and a Knowledge Graph within a single SQLite file. It enables agents to track relationships and context across sessions without relying on cloud services or external databases.

What it does

Kioku Lite fuses three search signals—BM25 keyword search, vector similarity, and Knowledge Graph queries—using Reciprocal Rank Fusion to provide structured memory. It operates fully offline, requires no Docker or cloud APIs, and integrates via simple pip installation.

Who it's for

It is designed for developers building AI agents that need persistent, structured memory across sessions, compatible with platforms like Claude Code, Cursor, Windsurf, and OpenClaw.

Why it matters

Kioku Lite addresses the limitation of existing memory tools that store only flat text or vectors by enabling causal reasoning and relationship tracking in AI agent memory without external dependencies.