Use cases
The question is always the same. The moment it gets asked is what changes.
Nobody wakes up wanting a trust score. They want to know whether to go ahead — and they want to know in the ten seconds before the money moves.
An agent about to pay
The reflexive pre-payment check.
An autonomous agent transacting on its own initiative has a wallet, no account, and no instinct. Before it signs anything to a counterparty it has never seen, one keyless call gives it a number it can gate on and a signal list it can log. The x402 rail exists for exactly this: no signup, no key, five cents.
x402 · POST /api/x402/check
Marketplaces and platforms
First-pass seller screening at signup.
Not every new seller can be reviewed by a human, and most do not need to be. Run the domain at onboarding, let the strong band through, and route weak and insufficient into the queue a person actually reads. The signals go in the audit trail, so a rejected applicant can be shown what was observed.
REST · one call per applicant
Finance and procurement
Before the first invoice from a new supplier.
A domain registered nine days ago, with no postal address and a certificate minted this morning, is worth a phone call before a wire — not because it proves anything, but because the five seconds are free and the wire is not. Payment-diversion fraud usually looks exactly like this.
Dashboard · check, then keep the record
Support and trust teams
Triage on a reported storefront.
When a customer reports a site, the first question is always the same set of facts. Getting them in one call, in a fixed shape, with dates and issuers spelled out, means the analyst starts from evidence instead of assembling it.
REST · scriptable, cheap enough to batch
Anyone with a card out
The advert followed you here. Should you buy?
Paste the domain. Domain age and the postal address answer most of it. A shop with a real trading address, policy pages and a three-year-old registration is a different proposition from one with none of those — and the product tells you which one you are looking at rather than telling you what to do.
Web · five cents, five seconds
Agent builders
A tool your agent can defend.
Every score comes with the arithmetic that produced it and a published model to check it against. When your agent explains why it declined a counterparty, it can cite a registration date and a certificate issuer rather than a vibe from a language model.
MCP · get_scoring_model is public
Where not to use it
Two jobs this is wrong for.
Publishing a score about someone
These numbers are a private first pass, not a rating you should put on a page next to a company’s name. A weak band mostly means new, and publishing it as though it meant otherwise would be unfair to a business that has done nothing wrong. Trust Check does not run a public directory of scored companies for the same reason.
Anything with a compliance obligation
No registry filings, no beneficial ownership, no sanctions screening. If a regulator expects Know Your Business, this is not it and will not stand in for it.
Trust Check reports publicly observable signals about a domain. It is not a certification, a credit check, or a fraud determination. A low score most often means a business is new or has a thin public footprint — not that it is dishonest. Read the signals and their evidence before acting on the number.