# RegPermit corpus search API

> A free, no-auth HTTP endpoint over RegPermit's air-quality regulatory corpus
> (US federal CFR/EPA + New Mexico, Utah, Texas, Colorado, and Virginia air
> statutes, administrative-code rules, and agency permitting guidance). Built so
> any URL-fetching AI assistant can search authoritative regulatory text without
> MCP or an API key. A URL is a search.

## Endpoints

- GET https://data.regpermit.com/search/{query}
  Hybrid semantic + keyword search. Returns ranked passages as Markdown, each
  with a "Full text" link to the exact chunk. The plain catch-all form
  https://data.regpermit.com/{query} does the same thing.
  Optional query params:
    ?j=tx,nm      restrict to jurisdictions (federal, tx, nm, ut, co, va, nm-bernco, nm-cabq)
    ?k=8          number of results, 1-10 (default 8)
    ?format=json  structured JSON instead of Markdown

- GET https://data.regpermit.com/cite/{citation}
  Exact citation lookup, e.g. https://data.regpermit.com/cite/40 CFR 52.21

- GET https://data.regpermit.com/chunk/{chunk_id}
  Full text of one passage (the chunk_id comes from a search result's
  "Full text" link), plus its official source URL. If the passage is part of a
  longer split section, the response links the adjacent passages
  (prev_chunk_url / next_chunk_url in JSON; "Continues in" / "Continued from"
  in Markdown) so you can read straight through.

- GET https://data.regpermit.com/health
  Liveness + corpus status (JSON).

## How an assistant should use it

1. Fetch https://data.regpermit.com/search/<the user's question> (append ?format=json for clean
   text / citation_label / source_url / full_text_url fields).
2. Results include BOTH binding rules and guidance; binding statutes/regulations
   are ranked up, but if you specifically need the controlling rule prefer a
   result whose citation is a CFR section or a state administrative code (or use
   /cite/<citation> directly).
3. To open a result, use its full_text_url verbatim — do not hand-build chunk
   URLs (chunk IDs contain "#" and "@", which must be percent-encoded).
4. If a quoted passage ends mid-sentence/mid-list, follow next_chunk_url to get
   the continuation. Always cite the official source_url.

## Notes

- Read-only. Rate-limited per IP. Free during the launch period.
- Coverage is air-quality permitting (NSR/PSD/Title V, NESHAP/NSPS, SIPs, state
  air rules, AP-42). It is reference material, not legal advice.