Every US government dataset. One API. Every response cited.
One key wraps SEC, FDA, Congress, NOAA, Census, FRED, court opinions, clinical trials, and 27+ more agent-ready categories. Every response ships with a machine-readable citation — source, license, retrieved-at, content hash.
$ curl https://api.katzilla.dev/agents/health/actions/fda-recalls \ -H "X-API-Key: kz_..." \ -d '{"classification":"Class I","_limit":1}' { "data": { "recalls": [{ "recall_number": "D-0412-2026", "product": "Metformin HCl 500 mg ER", "classification": "Class I", "reason": "NDMA impurity above ICH limit" }] }, "citation": { "source_name": "openFDA", "license": "CC0", "retrieved_at": "2026-04-23T14:08Z", "data_hash": "sha256:9c4e…" } }
Agencies, statistics offices, courts, and national open-data portals — not aggregators, not scraped blogs.
One API key. Every government source. Every response cited.
AI agents are only as trustworthy as their sources.
When an agent answers a legal question, a regulatory question, a clinical question — the user doesn't want a fluent-sounding hallucination. They want the primary source. Government data — SEC filings, Federal Register notices, FDA recalls, NIH clinical trials, Congressional records, court opinions, BLS labor stats — is the most authoritative, public-domain, citable corpus on earth. And it's almost entirely absent from how agents are wired today.
We wrap it. All of it. Behind one API key, structured for tool-use, with citation metadata baked into every response. So when your agent says something important, it can prove where the answer came from.
Around the API we ship the surround tooling agents need in production — Scrape for pages behind HTML, Signal for monitors + curated editorial digests, Ask for ad-hoc natural-language queries. Each optional, each colored its own way.