Measurement Log

What we recorded calling every API

This page is the raw output of our own testing, not a summary of anyone's marketing material. We send real HTTP requests to public API endpoints and record what comes back — status, latency, transport security, CORS headers and rate-limit headers. Nothing here is estimated, inferred or copied from a provider's documentation.

LAST RUN 2026-08-10

Latest run

271

endpoints called

244 ms

median response time

891 ms

90th percentile

257

responses captured

220

browser-callable (CORS)

235

negotiated TLS 1.3

41

published rate-limit headers

214

confirmed discontinued

1,212 of 1,426 catalogued APIs currently respond. 271 of those expose an endpoint we can call without credentials, so only those carry measured scores; the rest are scored from metadata and labelled as such on their pages.

Response time distribution

Median of three consecutive requests per endpoint, measured from a single vantage point. Your latency will differ by region — treat these as relative, not absolute.

under 100 ms
81
100–299 ms
88
300–599 ms
62
600–999 ms
15
1–2 s
20
over 2 s
5

Fastest endpoints we measured

Ranked by median response time across three requests.

API Category Response CORS TLS
VALR Cryptocurrency 27 ms no TLSv1.3
xkcd Games & Comics 29 ms no TLSv1.2
Nominatim Geocoding 29 ms yes TLSv1.3
FreeToGame Games & Comics 30 ms yes TLSv1.3
Drupal.org Open Source Projects 30 ms yes TLSv1.2
Final Space Video 30 ms yes TLSv1.3
AviationWeather Weather 30 ms no TLSv1.3
GrünstromIndex Environment 31 ms yes TLSv1.3
Chess.com Games & Comics 31 ms yes TLSv1.3
Dev.to Personality 31 ms yes TLSv1.2
Currency-api Currency Exchange 32 ms yes TLSv1.3
An API of Ice And Fire Video 32 ms yes TLSv1.3
Postali Geocoding 33 ms yes TLSv1.3
Coronavirus in the UK Health 33 ms yes TLSv1.3
Imgflip Entertainment 34 ms yes TLSv1.3

Transport security observed

TLS version actually negotiated during our requests.

TLSv1.3 235 endpoints
TLSv1.2 27 endpoints

Run history

Each weekly run calls every live endpoint again. Endpoints that stop answering drop out of the measured tier until they work again.

Run Endpoints tested Answered with data
2026-08-10 1213 271
2026-07-31 1213 252

How this is produced

  1. For each catalogued API we look for a callable endpoint — in its own code samples, in its documentation page, or at conventional paths on its own domain. Placeholder URLs and credentials are excluded.
  2. We send a GET request with a normal browser user agent and follow redirects. A response only counts as measured if it returns 2xx and valid JSON that is not an error envelope, a service banner, or an HTML page.
  3. We repeat the request three times and record the median and fastest round trip, plus the TLS protocol, certificate issuer, CORS header, rate-limit headers, content type and payload size.
  4. The response body is stored, truncated for readability, and shown on the API's page as a captured response.
  5. Endpoints that stop responding are re-checked; an API is only marked discontinued after two independent DNS resolvers confirm its host no longer exists.

Measurements come from a single network location and reflect one moment in time. They are a reproducible signal, not a service-level guarantee. See how API Score works for how these signals turn into a score.