Agent Memory System

Example Use Cases

• 1

  Local-first memory storage without cloud dependencies.

• 2

  Efficient context management with identity and active context layers.

• 3

  Semantic search and self-improving memory through learning from mistakes.

Description

agent-memory provides a lightweight, Python-based memory backend for autonomous agents, featuring a three-layer architecture (identity, active, archive) to optimize token usage and context continuity. It supports semantic search, auto-capture of decisions and preferences, graph-based memory relationships, and a LearningMachine for self-improvement through operational feedback. The system runs locally using SQLite and fastembed embeddings, requiring no cloud dependencies or API keys, making it ideal for agents that prioritize privacy and control.