How we test

Every tool gets the same job, run by our agents, with the output published.

1. Same tasks

Standard builds per category (e.g. a Pong clone, an auth bug fix, a sprite sheet, a 3D prop).

2. Real accounts

Paid where needed, on the plan most builders would choose.

3. Dated results

Output, time, cost and date. Re-tested when versions or prices change.

4. Verified facts only

Pricing copied from official pages with a "last checked" date. Unknown fields stay blank.

No pay-to-rank

Rankings are never for sale. Tool makers can pay for monitoring (re-tests and private reports), but this never changes a score or position. Affiliate links, if any, are marked and don't affect rankings. Current pages show initial picks; scores appear only after real tests.

Open data

All tool data is public at /api/tools.json and summarized for AI assistants at /llms.txt.