API & MCP server for AI agents

Why this exists

AI agents and answer engines increasingly need structured, provenance-bearing sources instead of scraping HTML. PermitBase's dataset — permit requirements, fees, codes adopted, and Census building-activity data — is already source-cited and versioned, so we publish it directly as a static contract any agent or developer can fetch, with the same grounding guarantee as the site itself: every response byte is either a verified field or an explicit, honest refusal. Nothing is ever summarized, derived, or guessed.

Available now — static JSON

• /data/v1/manifest.json — dataset version, schema version, coverage counts, and BPS activity vintage.
• /data/v1/coverage.json — every covered state, county, and jurisdiction, with permit-type counts. Check a jurisdiction is listed here before relying on it — anything absent is uncovered, never guessed.
• /data/v1/jurisdiction/{slug}.json — the full requirements record for a jurisdiction (department, codes adopted, local amendments, every permit type's requirements/documents/fees/timeline/ sources) fused with its Census building-activity slice. Requirements freshness (lastVerified) and activity freshness (bpsVintage) are always kept separate — never conflated.
• /llms.txt — a plain-text summary for AI crawlers and answer engines, linking the endpoints above.

No API key, no rate limit beyond ordinary fair use, no cost. This is the same free data as the website — just structured for machines instead of browsers.

Planned — MCP server

A remote MCP (Model Context Protocol) server at mcp.permitbase.app is built and locally verified, but not yet deployed. Once live, it will expose the same dataset as callable tools — coverage lookup, jurisdiction and permit detail, building activity, and jurisdiction comparison — so an AI agent can query PermitBase directly inside a conversation instead of fetching JSON by hand. It will remain retrieval-only: no model runs inside the server, so it structurally cannot hallucinate a fact.

Check back here or watch /llms.txt — it will be updated the moment the server goes live.

Important: this is a reference, not legal advice

Data served through these endpoints carries the same disclaimer as the site: it is a source-cited reference, not legal advice, and requirements change. Always confirm with the jurisdiction before submitting or relying on a figure for a real project.