api.foo
api.foo / Docs / How we test

How We Test: Methods, Price Sources and Update Policy for the api.foo Docs

Updated · By the api.foo docs team (how we test)
On this page

In one sentence: everything in the api.foo docs is tested before it is written. Tools run in an isolated environment and connect to the official APIs and api.foo with a fake key that cannot be charged, and we record the real responses. Prices come only from official pages and are dated.

How we test

  1. Isolated environment: tools are installed in a temporary folder and run with a temporary home or config directory, so no real configuration is touched.
  2. Fake key: we use a correctly formatted key that cannot be charged (for example sk-thisisafakekeyfortesting) against both the official API and api.foo, and record the status code and the exact error. An "invalid key" response shows the URL, format and auth method are right and only the key is wrong. It costs nothing and touches nobody's account.
  3. See what the tool really sends: we run a local server that only logs requests and point the tool at it, to see the actual URL, headers and model name. That is how we confirmed, for example, which model Claude Code requests by default and whether it appends /v1 itself.
  4. State the test environment: each guide lists the OS and tool version it was tested on, plus an "Updated" date. Newer versions may differ; the stated version is what we verified.
  5. Say what we did not test: if we could not verify something (say, whether an official endpoint is reachable from a given network), we write that instead of guessing.

Where prices come from

  • Only the model makers' official published prices, archived from the official pricing pages when we write and dated in the text ("as of October 2026").
  • Worked examples ("what does one task cost") are double-checked with a script.
  • api.foo's own prices are not written into articles; see the live model square.

Update policy

  • When official prices, major tool versions or error messages change, we re-test, update the article and its date.
  • Official links in articles are checked periodically and replaced when they move.

Disclosure

  • These docs are written by api.foo, an independently operated third-party API service (see What is api.foo) with no affiliation or partnership with Anthropic, OpenAI, Google or other model makers.
  • Guides show how to connect to api.foo and also how to use the official route; comparisons with other services use only their public information.
  • We do not publish ways around regional restrictions and never claim to be "official" or "authorized".

Found a mistake?

Email [email protected] with the page URL and what is wrong. Once verified we fix it and update the date.

FAQ

Are the configuration examples actually tested?

Yes. Before writing a guide we install the tool in an isolated temporary environment, configure it exactly as the guide says, and connect to the official API and to api.foo with a fake key that cannot be charged. We confirm the request reaches the right endpoint and gets the expected response (for example an invalid-key error), then note the test environment and version at the top of the article.

What does a fake key prove?

It proves the base URL, request format and auth method are right. The request reached the server and was understood, and was only rejected because the key is wrong. It costs nothing and touches no real account; the error messages quoted in our guides were captured this way.

Where do the prices come from?

Only from the model makers' official pricing pages, archived when we write and dated in the text. api.foo's own prices are not written into articles; the live model square is authoritative.

What if I find a mistake?

Email [email protected] with the page URL and what is wrong. Once verified we fix it and update the date at the top of the article.

Found a mistake? Email [email protected].