The Kin Score answers one question: how ready is this API to be discovered, understood, adopted, and operated — by a developer today and an agent tomorrow — without a sales call?
Every input is a public, machine-checkable signal. Does an OpenAPI exist at a fetchable URL. Is there JSON Schema. Are there documented plans and rate limits. Is there a security scheme declared. Is there a change log, a status page, a support channel, terms. Is there an MCP server or agent-facing artifact. No human quietly adds or removes points, which means the score is reproducible and you can check it yourself.
It carries a provenance dimension, which matters more than it sounds. My enrichment pipeline authors artifacts on providers’ behalf — modeled OpenAPI, candidate tool catalogs — and earlier versions of the rubric could not tell those apart from artifacts a provider actually published. That was a defect, not a nuance. Authorship now grades the credit: first-party publication counts fully, conformance to a standards body’s artifact counts less, and something I generated about you counts least.
The rubric is versioned in public, with a changelog, and it changes as it gets applied to new sectors. Scores move when the rubric moves, and that is stated rather than hidden.
