5 US States Open 12.4M Business Records via Free API

Accessing US company registration data is usually costly or blocked outright — OpenCorporates caps its free tier, and states like Delaware offer no bulk access at any price. Yet five states — New York, Colorado, Pennsylvania, Oregon, and Connecticut — publish their entire business registries through documented Socrata SODA APIs under public-domain or commercial-OK licenses, together totaling roughly 12.4 million entities with names, formation dates, addresses, and often registered-agent details.

To pull this data, the authors built a single config-driven Python script that maps each state's differently-named columns onto one unified schema and resumes cleanly after interruptions. The measured performance numbers defy the usual instincts: plain offset pagination is the fastest anonymous option at ~500 rows/sec with no meaningful slowdown at deep offsets, while keyset pagination and the CSV bulk-export endpoint are dramatically slower. Registering a free app token unlocks higher rate limits beyond this anonymous floor.

Several gotchas can silently corrupt naive pulls — some states store one row per officer or associated name rather than per entity, inflating counts 2-3x unless deduplicated, and CSV export headers don't always match API field names. States like California, Texas, and especially Delaware keep bulk data paywalled, while Florida and Ohio complicate access with flat-file formats or bot walls. The full verified dataset list, fetcher script, and findings are released as an MIT-licensed open-source Agent Skill, useful for engineers building company-matching, formation-trend, or business-survival analyses.