Why we built Bullion API

If you build fintech or e-commerce tooling, sooner or later you need a gold price. A checkout that shows jewellery prices in real terms. A dashboard for a pawnbroker. An invoice tool that denominates in grams. Nothing exotic — just "what is gold at, right now, in my user's currency."

The options were all bad

When we went looking for an API to answer that question, every option failed the same way. The free tiers were demo bait — a hundred requests a month, then a hard wall. The "affordable" tiers played rate-limit games: per-second throttles tuned so the advertised monthly quota is unreachable in practice. And the serious products wanted a sales call before they'd show you a price.

We didn't want a relationship with a sales team. We wanted a curl command.

What we built

Bullion API is one endpoint for gold spot prices, plus a timeseries endpoint for daily closes. Prices refresh hourly and are served from our own cache, so your request never waits on an upstream provider. Thirty currencies, ISO codes, X-API-Key header, done.

curl -H "X-API-Key: bullion_your_key" \
  "https://api.bullionapi.dev/v1/latest?currency=USD"

The response is plain JSON:

{
  "status": "success",
  "currency": "USD",
  "unit": "toz",
  "metals": { "gold": 4527.86 },
  "currencies": { "EUR": 0.9134, "GBP": 0.7783 },
  "timestamps": { "metal": "2026-07-08T11:58:50.781Z", "currency": "2026-07-08T11:58:50.781Z" }
}

Quota is predictable by design. Every response carries x-ratelimit-limit, x-ratelimit-used and x-ratelimit-remaining headers, the counter resets on the first of the month, and when you hit the cap you get a 429 with a retry-after — never a surprise invoice, because we don't charge overage fees.

The free tier is honest

The Free plan is 30 requests a month. That is not enough to run production on, and we're not going to pretend it is. It's enough to evaluate the API properly: wire it into your app, run it for a few weeks, watch the uptime, and decide. If it works for you, Pro is $9 a month for 50,000 requests. If it doesn't, you've lost nothing.

Built for agents as much as humans

A growing share of "developers" integrating APIs are AI agents reading docs on a user's behalf. So the site ships machine-readable versions of itself: /llms.txt and /llms-full.txt for agent crawlers, /openapi.json (a full OpenAPI 3.1 spec) for function-calling clients and code-gen, and markdown mirrors of the docs, pricing, terms and privacy pages at *.md URLs.

What's next

Gold-only at launch was a deliberate scope cut — one metal, done properly, beats five done loosely. From here, we ship what people ask for, in the order they ask for it, and we don't announce features before they're running in production. Everything on this site works today or it isn't on this site.

If there's something that would change what you build — a metal, a currency, an endpoint shape, an integration — tell us. A real person reads every request, and future blog posts will share what we learn next.