Waggle Memory

Graph-backed persistent memory for AI agents using typed nodes and relationship edges stored in SQLite.

Waggle Memory provides AI agents with persistent, structured memory using a knowledge graph stored in SQLite. Rather than flat note storage, it creates typed nodes (facts, decisions, preferences) and relationship edges that persist across conversation sessions. Tools support automatic fact extraction from conversations, semantic graph queries, temporal diff tracking, and context priming for new sessions. Benchmarked at 97.4% R@5 accuracy on LongMemEval with 91 passing tests and sub-2ms query times.