Know Your Agent (KYA) providers: what each one checks

KYC asks who a customer is. KYA asks which agent is acting, who it acts for, what it may do, and whether you can rely on the answer.

Skyfire describes KYA as "verifying an AI agent’s identity, binding it to a responsible person or organization", then deciding what the agent is authorised to do. Its page is plain about the edges: "What it does not replace: KYC, OAuth, IAM, API keys, or payment-network infrastructure."

There is no single standard. Persona's page says "There isn't a singular way to perform KYA", and the providers below use the same three letters for different checks: some verify the human behind an agent, some verify the business that built it, some sign a credential the agent carries, and some keep a record of what it did. An open specification, KYA-OS, is being developed at the Decentralized Identity Foundation.

KYA and KYC are not either-or. KYC verifies a person or a business, usually with documents, a selfie or company registers. KYA verifies the software acting for them, and its authority. Many KYA products start with KYC and then bind the agent to the person they verified.

Five questions. Ask every provider, APIR included, which of them it answers, and how you would check the answer yourself.

• Whose agent is it? The person or organisation that answers for it. Some providers verify that human with a government ID and a live selfie, some verify the developer's business, and some record the principal an operator names and sign it.
• What may it do? The scope: allowed actions, a spending cap, an expiry, and who granted it. When one agent delegates to another, the chain should lead back to a person, and every link should still hold.
• What has it done? What happened after the agent was let in. Some keep audit logs for the platform that runs the check, one aggregates behaviour reported by operators, and some keep a hash-chained record in which a later change shows.
• Can you check it without trusting the provider? A signature you can verify yourself against a published key is worth more than a green tick on their dashboard. Ask which parts you can check yourself and which parts are still the provider's word.
• What does it never prove? No KYA check proves an agent will behave well tomorrow. Trulioo says so on its own page, and the same is true of APIR.

For when you must know the real person or business behind an agent. Most bind the agent to someone they verified with documents, a selfie or business checks, and several also watch the agent's traffic.

Pick one of these when a rule or a fraud risk means you must know the real human or business behind an agent, or when you need to spot and police agent traffic on your own site. APIR does neither: it does not run document or biometric checks on anyone, and it does not sit in front of your website.

A credential the agent carries from service to service: who stands behind it and what it may do, signed so the receiver can check it.

Pick one of these when you build agents and want them to carry a portable, signed credential that other services accept. A credential answers who stands behind the agent and what it may do at the moment it is presented; a history of what the agent did afterwards is the job of Kind 4.

For when the agent pays. The networks recognise approved agents at checkout and tie the payment to the consumer's permission.

Pick one of these when an agent spends money on a card. APIR is not a payments network, issues no payment tokens and moves no money.

The other kinds mostly check an agent at the door, and some keep logs for the platform that ran the check. These keep a history of the agent that someone else can look up.

The two are different records. KnowThat.ai aggregates what operators report about an agent into a trust score. APIR keeps what the agent and its owner put on the record about the agent's own work, chained so that a changed or deleted entry shows, and its trust score comes only from the Trust Genome, computed from that record and its papers: an agent without enough evidence is shown Unrated, never given a made-up number.

APIR gives every AI agent signed papers and a tamper-evident work history that buyers, auditors and insurers can check without trusting you.

Most of the checks above centre on the door: who is behind the agent, and is it allowed in. APIR answers part of that: a signed identity, and a mandate naming the principal its operator gives, without verifying that person. It adds the part a buyer, a platform or an insurer asks next: what has this agent done, on a record where any later change shows? A public passport's signature checks without trusting APIR; the log book's chain check and its timestamp tokens still rest with APIR today.

1. 1. A person checks one agent, free

   Paste an agent's name, its passport ID or its agent ID at apir.ai/check-agent. No account, never paywalled. The passport's signature is re-checked against the published key, the mandate is read back, and the log book's events are counted with the time of the last one. An agent with no passport returns NO_RECORD: a passport is issued at a score of 70 or more, and is found by name only once public. The registry lists the agents whose owners chose to list them publicly; any public passport opens at apir.ai/verify-passport by its passport ID.

2. 2. A platform checks at volume

   Call the Resolution API with an API key from apir.ai/developers. It takes an agent name (public credentials only), a credential ID or an agent ID, and returns a verdict (VERIFIED, EXPIRED, REVOKED, UNSIGNED, UNCHECKED, SUPERSEDED or NO_RECORD) with identity, authority and record. The first 1,000 checks each month are free, then each answered check is metered; the price is on the developers page.

   GET https://apir.ai/api/v1/resolve?q=<agent name, TP- credential id or agent id>
   Authorization: Bearer <your API key>

3. 3. Check the signature yourself

   For a public passport, fetch its envelope from https://apir.ai/api/verify-passport?id=<passport id> and the keys at apir.ai/.well-known/jwks.json, published under did:web:apir.ai. Ask the owner for its organisation ID (signed, but not printed in the envelope), rebuild the signed bytes from the canonical payload in the specification, and verify them with any Ed25519 library; how to verify an AI agent has the script. The check and resolve answers carry no signature bytes, and the record count and the chain check are APIR's answer. An AI agent can run the same lookup through APIR's read-only MCP server at https://apir.ai/api/mcp, listed in the MCP registry as ai.apir/verify.

NO_RECORD means nothing is on file. It is not evidence of wrongdoing; it means any claim the agent makes about itself is unverified. Every verdict carries the moment it was made: true as of that moment, never forever. To put your own agent on the record, see how to prove what your AI agent did.

Name the limits first. A KYA answer you cannot rely on is worse than none.

• It does not verify the human behind an agent with document or biometric KYC. The mandate records the principal an operator names; if you must know who that person really is, use a KYC provider from Kind 1.
• It does not detect, challenge or block bot traffic on your website. Vouched Agent Checkpoint and Sumsub describe doing that, and Incode describes detecting and classifying agent activity.
• It is not a payments network. It issues no payment tokens and moves no money.
• It certifies no one, and it is not an insurer, a broker or a compliance platform.
• The record shows what was put on it. The log book's entries are what the agent and its owner recorded, whether the agent runs on APIR or anywhere else; it proves what was recorded and shows any later change, not that the record is complete. A count of events is not a judgement of quality.
• It has no outside witnesses countersigning its chain heads and publishes them to no transparency log. Its nightly RFC 3161 timestamp tokens are public only for agents with a public passport, through the verify-black-box endpoint, and inside an owner's export. The passport, the check, the resolve answer and the Evidence Pack do not carry them yet.
• The log book's chain check runs on APIR's side, for public passports. An agent's events are not public: to recompute them offline, an outsider needs the free export from its owner.
• The mandate's signature bytes and its earlier versions are not yet published, and an owner can replace a mandate.
• It is not a reputation network: it does not collect reviews of an agent from other operators.
• Only public credentials resolve by name. A private one resolves only for a caller who already holds its exact ID.

We don't promise results. We prove whatever happens.

They stack. Most platforms that need one kind of KYA will need two.

• You must know the real person behind an agent, for KYC rules or fraud: an identity verification firm such as Persona, Sumsub, Incode, Microblink or Vouched. If it is the developer's business you need verified, Trulioo checks it through KYB and AML.
• You need to spot and police agent traffic on your own site: Vouched Agent Checkpoint, which detects agent traffic and can allow, block, challenge or redirect it, or Sumsub, which detects automation and challenges it with a liveness check rather than blocking. Incode also detects and classifies agent activity.
• The agent pays: a card network programme such as Visa Trusted Agent Protocol or Mastercard Agent Pay, or Skyfire's KYAPay tokens.
• You build agents and want a portable credential on an open standard: KYA-OS, W3C Verifiable Credentials, or Dock Labs' Truvera Agent ID.
• You want to know how other operators found an agent: KnowThat.ai, the reputation registry powered by Vouched.
• A buyer, a platform or an insurer needs to check an agent's papers, mandate and record without trusting the vendor: that is the job APIR does. Check an agent, free, or resolve agents at volume with the first 1,000 checks a month free.

A marketplace can bind an agent to a verified human with a KYC provider, accept its card network token at checkout, and still resolve its APIR papers and record before it lets the agent act. Agent identity inside one company is compared in AI agent audit and compliance tools compared.

What is Know Your Agent (KYA)?

Know Your Agent (KYA) is checking an AI agent before you let it act: whose agent it is, what it may do, and whether you can rely on the answer. It borrows from Know Your Customer (KYC) but checks the software acting for a person or business, and its authority, not only the person. There is no single standard yet: providers' own pages, read 1 November 2026, describe different checks under the same name, and an open specification, KYA-OS, is being developed at the Decentralized Identity Foundation.

Who provides Know Your Agent (KYA)?

Four kinds of provider, read from their own pages on 1 November 2026. Identity verification firms extending KYC to agents: Vouched, Trulioo, Persona, Sumsub, Microblink and Incode. Agent credentials and open specifications: Skyfire's KYA tokens, Dock Labs' Truvera Agent ID, KYA-OS at the Decentralized Identity Foundation and W3C Verifiable Credentials. Card networks' agent programmes: Visa Trusted Agent Protocol and Mastercard Agent Pay. Record-based checks: KnowThat.ai, a reputation registry powered by Vouched, and APIR, which returns an agent's signed identity, its human mandate and a count of its hash-chained log book in one answer.

How is KYA different from KYC?

KYC verifies a person or a business, usually with documents, a selfie or company registers. KYA verifies the AI agent acting for them: which agent it is, who it acts for, and what it is allowed to do. Many KYA products start with KYC and then bind the agent to the person they verified. Skyfire's page says KYA does not replace KYC, and Trulioo verifies the developer organisation through KYB and AML checks before it profiles the agent. A workflow can need both.

Can I verify an AI agent for free?

Yes, for an agent on APIR. Paste its name, its passport ID or its agent ID at apir.ai/check-agent: no account, never paywalled. The passport's signature is re-checked against the published key, the mandate is read back and the log book's events are counted. An agent with no passport returns no record (a passport is issued at a score of 70 or more, and is found by name only once public), which is not evidence of wrongdoing; it means its claims about itself are unverified. KnowThat.ai also offers a lookup of an agent's trust score on its own site, with no price stated on the page read; most other providers on this page sell their checks to businesses, so check their pages for any free tier.

What does a KYA check prove?

Only what the provider checked, as of the moment it checked. Depending on the provider that can be a verified human behind the agent, a verified developer business, a signed credential with a scope and an expiry, or a record of past activity. None of them proves the agent will behave well: Trulioo's own page says verification does not guarantee an agent's future behaviour. APIR shows its passport was signed by the key published under did:web:apir.ai and, for a public passport, that APIR's recompute of its log book finds no change since each entry was written; the log book shows what the agent and its owner recorded, not that the record is complete.

How do platforms check agents at volume?

By API, inside the request. Under Visa's Trusted Agent Protocol the merchant checks each signed agent request against the agent's public key; Mastercard Agent Pay uses registered agents and Agentic Tokens; Skyfire's tokens carry identity claims to a seller, and Trulioo gives counterparties API and MCP access to an agent's profile. APIR's Resolution API, GET https://apir.ai/api/v1/resolve with an API key, answers identity, mandate and record in one call; the first 1,000 checks each month are free, then each answered check is metered, priced at apir.ai/developers. A read-only MCP server at https://apir.ai/api/mcp lets an AI agent run the same check, free.

Is APIR a KYC provider?

No. APIR does not verify the human behind an agent with documents or biometrics, does not detect bot traffic, is not a payments network and certifies no one. It keeps the agent's signed papers, its mandate and its tamper-evident log book, and lets anyone check them. If you must know the real person behind an agent, use a KYC provider; APIR sits beside it.