About MechRegistry

MechRegistry is a registry of Mechs. A Mech is an AI agent-curated knowledge base that keeps one validated, ontology-grounded, evidence-backed record per entity, with a LinkML schema and an append-only curation history. The canonical Mech is DisMech, the Disorder Mechanisms Knowledge Base. What is a Mech? says more.

The registry records, for each Mech:

  • what one record represents, and how many records there are
  • how records are identified and which ontologies ground them
  • which upstream sources the records are seeded or harmonized from
  • the curation model: which agents write it, what human review gates it, what evidence a claim must carry, and what validation runs in CI
  • its products: record corpus, schema, browser, exports and documentation
  • its licenses, contacts and publications
  • typed relations to other Mechs

How it is built

Every Mech is one Markdown file at mech/<id>/<id>.md. Its YAML front matter is a Mech object in the MechRegistry LinkML schema. A build step validates every entry against the schema with closed validation, so a misspelled field or an unknown enum value fails the build, and then concatenates the entries into the registry files the site and other tools read:

The site is a Jekyll site on GitHub Pages. The design follows KG-Registry: Markdown with YAML headers, one LinkML model, static pages, no server.

Licensing

Registry data are released under the CC0 1.0 Universal waiver. The repository code is released under the BSD 3-Clause license. Each Mech carries its own license, shown on its page.

Who maintains it

MechRegistry is maintained by the Monarch Initiative. Source and issues are on GitHub.