# AI Systems Atlas

> A directory of reviewed agent, memory, and assistant systems, provider-independent model releases and the labs that develop them, interoperability specifications, managed inference services, self-operated local runtimes, unscored agent packs a host agent installs, and unscored robots recorded from what their makers document, plus a complete attributed models.dev source snapshot. Imported source metadata remains distinct from human conclusions.

## Data

Published JSON, no auth, no rate limit — fetch directly. Atlas-authored data is licensed under CC BY 4.0:
reuse it, including commercially, with credit to AI Systems Atlas
(https://peacefulcoexistance.com/) and a link to the licence.

- [projects.json](https://peacefulcoexistance.com/projects.json): reviewed agent, memory, and assistant systems with editorial scores
- [taxonomy.json](https://peacefulcoexistance.com/taxonomy.json): enum values, license identifiers, and score-profile definitions used by every other file
- [exclusions.json](https://peacefulcoexistance.com/exclusions.json): reviewed scope-boundary decisions — why a candidate was not included
- [license-evidence.json](https://peacefulcoexistance.com/license-evidence.json): scoped, reviewed license and terms evidence per project
- [specifications.json](https://peacefulcoexistance.com/specifications.json): unscored interoperability protocols, metadata schemas, instruction conventions, and package formats
- [inference-services.json](https://peacefulcoexistance.com/inference-services.json): managed inference services with a dedicated operational-service score
- [local-runtimes.json](https://peacefulcoexistance.com/local-runtimes.json): self-operated inference runtimes with a dedicated runtime score
- [models.json](https://peacefulcoexistance.com/models.json): reviewed provider-independent model releases with a dedicated access-and-deployability score and, once models.dev lists the release, models.dev source metadata
- [models-dev.json](https://peacefulcoexistance.com/models-dev.json): complete commit-pinned models.dev source catalog, attributed under its MIT License and carrying no Atlas review or score
- [packs.json](https://peacefulcoexistance.com/packs.json): unscored agent packs a host agent installs, recorded for what each one installs, its hosts, install mechanism, and licence
- [labs.json](https://peacefulcoexistance.com/labs.json): unscored AI labs that develop the reviewed model releases, with headquarters, type, where each publishes, and the names the other files use for it
- [robots.json](https://peacefulcoexistance.com/robots.json): unscored robots whose maker's own documentation names a learned model or policy, or documents a way to run your own, recorded for hardware, developer access, availability, and terms as the maker documents them, never what a robot does

## Reference

- [Taxonomy](https://github.com/katagun/ai-systems-atlas/blob/main/docs/TAXONOMY.md): family, role, and trait definitions behind every enum
- [Curation](https://github.com/katagun/ai-systems-atlas/blob/main/docs/CURATION.md): the review workflow and inclusion rules applied to every record
- [Data model](https://github.com/katagun/ai-systems-atlas/blob/main/docs/DATA_MODEL.md): field-by-field schema for the published files
- [Agent Skill](https://github.com/katagun/ai-systems-atlas/blob/main/skills/ai-systems-atlas/SKILL.md): a packaged skill for querying this data, in the Agent Skills format