PRODUCT

Know Your Destination.

Payment systems verify who is paying. Nobody verifies who is getting paid. Know Your Destination is that missing check, and Kleyr runs it before the money moves.

kyd · product demo

01The counterpart nobody built

Payments already know how to interrogate the payer. KYC identifies the customer. Know Your Agent is arriving to identify the software acting for them. Both answer the same question: who is sending the money?

Nothing answers the other one. Where is it going?

A human buyer answered that informally — a brand they recognised, a domain that looked right, a checkout that felt normal. An agent has none of that. It sees a URL and a price. It cannot hesitate, and it will pay a storefront registered four hours ago.

KYD is the destination side of the transaction: the merchant, the domain, and the endpoint the money reaches, verified before the payment, not reconciled after it.

02How it works

  1. 01

    Agent reaches a destination

    A checkout it has never transacted with.

  2. 02

    It asks Kleyr

    One POST to /v1/verify, carrying the destination URL.

  3. 03

    GO / NO-GO

    Four scores, one verdict, one audit_id. Before any money moves.

03Four score layers

The verdict is not one opinion. Four independent layers score the destination, each built from encoded expert knowledge, and each able to sink a transaction on its own.

01

Security

Whether the destination is technically what it claims to be: domains, certificates, phishing infrastructure, brand impersonation.

02

Compliance

Whether the transaction is lawful where it lands: jurisdictions, product roles, sanctions, licensing.

03

Integrity

Whether the merchant plays fair with a machine buyer: prompt injection, manipulation, deceptive checkout flows.

04

Public Trust

Whether the rest of the world agrees: registries, reputations, reviews, track record.

04The verdict, and the evidence under it

Kleyr returns four scores, one verdict, and an audit_id. The verdict is the operational answer: pay, or do not. The audit reference is what matters afterwards — a timestamped, minimised, exportable record of what was known about that destination at the moment of the decision.

Not a log line. Evidence.

api.kleyr.ai → agent 
{
  "audit_id": "kly_8f3c1d0a7b2e",
  "url": "https://shop-nova42.com",
  "scores": {
    "security": 91,
    "compliance": 88,
    "integrity": 84,
    "public_trust": 79
  },
  "overall": 83,
  "verdict": "GO"
}

05Built for integration

KYD is live as an API. One request carries the destination, and the verdict comes back in the same envelope whether you wait for it or stream the run as it happens. Keys are issued to design partners.

agent → api.kleyr.ai 
curl -X POST https://api.kleyr.ai/v1/verify \
  -H "X-API-Key: $KLEYR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "url": "backmarket.com", "buyer_country": "FR" }'

The quickstart, the endpoints and the verdict thresholds are on the developers page; the full reference is generated from the OpenAPI spec at api.kleyr.ai/docs.

06Who runs the check

Any link in the chain between an agent and a merchant. Vertical AI agents transacting on sites nobody has vetted. The frameworks they are built on, shipping the check as a native primitive. Payment service providers and rails screening agent-initiated payments before settlement. Banks that want to qualify those payments before they leave the account.

Kleyr is neutral by structure: not the agent, not the merchant, not the rail. A verdict only holds if the referee has no stake in the game.

The standard for agentic trust is being written. Write it with us.

Become a design partner → Try the demo