Every measurement this site renders, as JSON, from 5 calls that need no key.

Every kit with its score, price, licence, stack and last commit. Filterable and sortable on the same axes the directory is.

stack

one stack slug, e.g. nextjs, laravel, prisma

ecosystem

Next.js, Laravel, Rails, Django, Node.js, Wasp

pricing

free or paid

evidence

hands_on or documented

max_price

an integer in USD

sort

score (default), name, price, updated

limit / offset

limit caps at 200

One kit in full: every capability with the file or page it was found in, every price point with its source, and the install transcript where we ran it.

Two to four kits side by side, with the fourteen capabilities resolved per kit.

slugs

two to four slugs, comma separated

The scoring method itself: the components, their point ceilings, and how each one is computed. This is the same constant the score is calculated from, so it cannot drift from the arithmetic.

The counts on the landing page, and when the directory was last refreshed.

The one thing this site publishes that is meant to live on somebody else’s page. It states this index’s own measurement of that kit, in words, and it links back here. The value is not settable by the embedder and there is no ?message= parameter: a badge whose number the holder controls is a sticker, and one forged badge ends the usefulness of every honest one. Add ?theme=light for a light card. A maker who wants the listing to record that a maker has claimed it does that on the claim page.

It carries measurements, not opinions. There is no recommended field, no editor’s pick and no ranking that is not the score arithmetic applied to the numbers in the same response. Where a component could not be measured, a kit with no public repository has no commit history to read, the points are simply not awarded and the coverage field says how much of the method was applicable at all. Read how the score is built before treating a total as comparable across kits with different coverage.