Where to get the data: stocks, weather, books, maps and more

The instinct when you need data is to scrape the site that displays it. Often that's the worst available option: slower, more fragile, legally murkier, and it breaks the first time someone changes a CSS class.

Work down this order instead. An official API. Then a bulk download or open dataset. Then the site's own undocumented JSON endpoint. Then scraping the HTML. Most domains in this guide are solved at step one or two.

Free tiers and terms in this area change frequently and some have been cut hard in recent years. Everything here was checked in September 2026 — verify the current limits before you build on them.

Stock and financial data

The hardest category to get free, because the data has genuine commercial value and exchanges licence it aggressively. Expect meaningful limits.

Realistic advice: for anything beyond a hobby, price data is the one category where paying is usually correct. The free tiers are designed to be just enough to prototype.

Weather

The best-served category on this list, and one where you genuinely don't need to pay.

Weather is also the ideal first data bot: no key needed with Open-Meteo, a clean JSON response, and a genuinely useful output.

Books and publications

Note that metadata about books is largely open while the text mostly isn't. Plan around that distinction early.

Maps, places and geocoding

The category with the strictest etiquette, because the free infrastructure is donated and easy to abuse.

Government and public data

Consistently the most underused source, and often the only place a given fact exists officially.

That last row is worth dwelling on. Municipal planning and licensing data is public, structured enough to parse, updated constantly, and almost entirely unexploited — which makes it a strong candidate if you're after a dataset nobody else has.

Everything else, briefly

Four habits that save you weeks

Cache everything, immediately. With Alpha Vantage at 25 calls a day you'll exhaust the quota debugging before you've built anything. Write every response to disk on first fetch and read from there while developing. This also makes you a good citizen of the free APIs.

Store the raw response, not just the parsed fields. When you later want a field you didn't extract, you have it. When the format changes, you can compare. Disk is cheaper than re-fetching a year of history you can't re-fetch.

Read the licence before you build, not after. "Free to call" and "free to republish" are different claims. Open-Meteo's free tier is non-commercial. Google Maps restricts storing results. OSM requires attribution. None of this is onerous, but all of it is easier to honour at the design stage than after launch.

Prefer the bulk download. If a source offers a dump, take the dump. It's faster, it has no rate limits, it doesn't burden anyone's server, and it lets you query in ways an API never would.

A first data bot, in one line of thinking

Pick a source with no key — Open-Meteo, Open Library, a government CSV — fetch it on a schedule, store every response, and let the history accumulate. That's the change watcher with a JSON endpoint instead of an HTML page, and it's the beginning of every dataset worth having.