I built an open-source cognitive memory layer for AI agents in Go — because amnesia shouldn't be a feature

A developer released Stash, an open-source Go-based memory layer for AI agents.

Stash stores experiences, goals, and failures, then synthesizes them into a knowledge graph.

It gives agents persistent memory across sessions, enabling better long-running reasoning.

The project is self-hosted and aims to reduce reliance on retraining or prompt stuffing.