API rate limits

Data.gov's aviation and cargo endpoints meter DEMO_KEY at 30 requests per IP per hour and 50 per day, against 1,000 per hour on personal keys. Apple's iTunes Search API permits roughly 20 calls per minute, with heavier use requiring the Enterprise Partner Feed. GitHub allows 60 requests per hour unauthenticated, 5,000 authenticated, but only 10 search requests per hour.

Limits also price the ladder. Alpha Vantage's free tier allows 25 API requests per day; $49.99-$249.99 per month buys 75-1,200 requests per minute with no daily limits on premium plans. Nasdaq Data Link returns only limited or sample data without a key. Where no API exists, politeness rules stand in: the Data.gov software datasets catalog requests a 10-second robots.txt crawl delay.

Limits convert 'how big is the source' into 'how long collection takes': a walk needing 3,000 requests takes 100 hours on DEMO_KEY, 3 hours on a personal key.

• Sampling masquerades as coverage. A vendor capped at iTunes Search's roughly 20 calls per minute may deliver a storefront slice and call it a market view; heavier use requires Apple's Enterprise Partner Feed.
• Paid ladders are the honest tier system. Alpha Vantage's 25 requests per day versus $49.99-$249.99 per month for 75-1,200 requests per minute shows exactly what money buys.
• Politeness directives are binding signals. A requested 10-second crawl delay is the robots.txt equivalent of a rate limit, and ignoring it risks the blocking behaviors cataloged elsewhere.

Use these checks to test whether a source delivers what the concept promises:

1. Register the free key and re-measure. GitHub's jump from 60 to 5,000 requests per hour authenticated changes feasible harvest size more than any tooling tweak.

1. Price the tier ladder openly. Alpha Vantage's steps from 25 requests per day to $49.99-$249.99 per month should appear in your cost model, not be discovered at renewal.

1. Check whether the SDK is crippled without a key. Nasdaq Data Link returns only sample data without one - demos prove little.

1. Respect crawl-delay directives. A 10-second robots.txt delay implies about 8,640 requests per day - plan around it or negotiate.

1. Favor bulk files where limits bind. Bulk download leads all access methods in this catalog (725 of 1,744 datasets) precisely because it escapes per-request metering.

Where to see it in context: air-freight-logistics data, application-software data.

Terms that travel with rate-limit planning:

• api-key-registration
• api data gov key rate limits
• cursor-pagination
• web scraping access
• automated collection prohibition
• web catalog scraping