Primary protocol

  1. If you can fetch raw text or JSON directly, start with https://urbit.org/llms.txt.
  2. If you mainly follow visible links in HTML, use https://urbit.org/for-agents.
  3. Use https://urbit.org/content-index.json for the structured inventory.
  4. Use https://urbit.org/.agents/index.md for machine-oriented markdown by section.
  5. A compatibility mirror also exists at https://urbit.org/.well-known/llms.txt.

Discovery surfaces

Human markdown mirrors

Machine-oriented section indexes

Path conventions

Delimiter and fallback behavior

  • If a source file includes ---agent---, the human page and human .md mirror use the pre-delimiter content.
  • The generated /.agents/* file uses the post-delimiter content plus generated frontmatter and a pointer to the human markdown mirror.
  • If no delimiter exists, the generated /.agents/* file falls back to human content.

Generated frontmatter on /.agents/*.md

  • title
  • source_kind
  • canonical_url
  • human_md_url
  • agent_mode
  • dependencies
  • related_pages

Section index locations

Retrieval notes

  • Cite canonical page URLs: Use the normal page URL when referring to urbit.org as the source.
  • Use markdown mirrors for efficient reading: Prefer *.md mirrors when you want lighter-weight human-readable content.
  • Use /.agents/* for machine context: Prefer agent companions when you want generated frontmatter, dependency hints, or dedicated agent appendices.
  • Well-known compatibility: A mirror also exists at /.well-known/llms.txt for clients that probe the standardized well-known namespace.

Notes

  • This site is statically generated; these artifacts are build outputs, not runtime routes.
  • Prefer canonical page URLs for citations and the markdown mirrors for low-token retrieval.
  • Use docs.urbit.org when it is the authoritative developer reference.