agent-memory-atlas

What the atlas is

Agent Memory Atlas is a field guide to memory systems for AI agents. It compares memory units, storage models, write paths, retrieval mechanics, correction semantics, trust models, agent integrations, and operational risks across systems whose source is publicly readable. The project includes some systems that are not open source and names them rather than averaging over them, and every report links to the exact Git commit that was analyzed so the analysis is reproducible.

How it is organized

The repository separates individual system reports under content/systems, reusable architecture pattern guides under content/patterns, a cross system comparative overview, and the review and synthesis methodology. A site directory holds the designed homepage and a Discord invite redirect, templates hold the shared document page template, and docs hold the generated static site published by GitHub Pages. The layout is meant to be both human readable and machine navigable.

Contributing and tooling

An AGENTS.md file tells an AI agent what to read to build memory for a product and in what order, and skills under .agents handle screening a checkout for auto executing hooks and unpinned dependencies, researching a new system, and designing memory from the atlas. The build requires Pandoc and runs with npm run build and npm run serve, then serves locally on port 4173. The code is MIT licensed and the snapshot shows 44 stars with 4 forks.

Editorial conclusion

The project is MIT licensed and reports 44 stars with 4 forks as of August 2026. It publishes a static site comparing agent memory systems, with each report pinned to the specific source commit it analyzed.