Show HN: FERNme – agent memory that updates with ~zero LLM calls

Category: library

Tags: agent-memory, graph-based, user-owned, privacy, cost-efficiency, hebbian-learning, memory-layer

Score: 7.3/10 (Innovation: 8, Technical: 8, Documentation: 8, Utility: 5)

FERNme is a zero-LLM-call agent memory layer that uses a sparse Hebbian graph to update and retrieve user profiles, with flat token cost and interpretable design. It learns from behavior and outcomes across domains, enabling user-owned, cross-site memory with privacy guarantees and cost-quality dials.

Target audience: backend devs, ai-engineers, data engineers

Repository: https://github.com/mirkofr/FERNme · Python · NOASSERTION · 1 stars

View on Hacker News