Agent Memory Atlas

A field guide that maps AI-agent memory systems for humans and agents alike: structured source records, taxonomy axes, tags, confidence scores, provenance, and verification hints - an atlas rather than another paper list.

View source on GitHub

Key takeaways

• 01

  Taxonomy axes (store type, write policy, recall policy) organize the space

• 02

  Confidence scores and provenance belong in memory metadata

• 03

  Verification hints make memory auditable

Flows built on this research

Memory & Context

Episodic, Semantic, and Procedural Stores

Organize agent memory by kind - what happened, what is true, and how to do things - instead of one undifferentiated pile.

4 steps · 60-90 minutes