Know Your Agent (KYA) guide

What is Know Your Agent (KYA)? A guide to verifying AI agents

Know Your Agent (KYA) is a risk-based framework for establishing and maintaining trust in AI agents. It involves defining their identity, binding them to responsible human or organizational principals, and enforcing policy and auditability across all of their actions. In practice, before an agent acts, KYA settles who the agent is, which verified human or business stands behind it, and what it is allowed to do.

Jovan Jovanovic · Chief Technology Officer, Incode Published November 2026

Know Your Agent (KYA) at a glance

Stands for: Know Your Agent
Verifies: An AI agent and the human or business accountable for it
Extends: Know Your Customer (KYC) and Know Your Business (KYB)
Core stages: Identify, bind, scope, monitor, and step up, revoke, or audit
Where it matters: Agentic payments, customer-facing agents, and regulated industries
Standards status: Emerging; payment networks announced an exploratory KYA collaboration in September 2026
Related terms: Agentic fraud, Know Your Human (KYH), agentic identity

On this page

01What is KYA?
02Why does KYA matter?
03How does KYA work?
04What three questions does KYA answer?
05How is KYA different from KYC and KYB?
06How does KYA compare with OAuth, API keys, and agent authentication?
07Is all AI agent traffic fraud?
08Which KYA frameworks and standards exist today?
09Where is KYA used?
10How should you evaluate a KYA approach?
11How does Incode Know Your Agent verify AI agents?
12Frequently asked questions

What is Know Your Agent (KYA)?

Know Your Agent (KYA)

Know Your Agent (KYA) verifies an AI agent and ties it to the person or business accountable for it, so a platform knows who is behind the agent before it acts.

The idea borrows from Know Your Customer (KYC), the checks banks and regulated platforms run to verify a person before opening an account. KYC asks whether a person is who they claim to be. KYA asks a harder question, one that has emerged as more users rely on agents to act on their behalf: when software takes an action, who is it for, and did that person truly authorize the action?

A person or a business can sign a contract, hold a bank account, and answer to a regulator. An AI agent holds only the authority that a person or business delegates to it. That is why every KYA approach, whichever network or platform defines it, holds that an agent is only as trustworthy as the verified human or organization it is bound to.

In identity and payments, KYA means verifying AI agents, a separate practice from the “know your agent” checks that vet real estate or insurance agents. It also covers different ground from workforce identity and access management (IAM). IAM governs the agents a company deploys itself, while KYA governs the agents that arrive at its front door acting for customers.

Why does Know Your Agent (KYA) matter?

Know Your Agent (KYA) matters because AI agents have started to act on their own for the people who send them, and every platform they reach has to decide whether to let them in or not. One decision represents the potential for fraud, and the other represents lost revenue.

Agents are moving from recommending to doing. They book travel, fill carts, open accounts, pay bills, and call APIs without a person clicking through each step.

By 2030, AI agents could orchestrate $900 billion to $1 trillion of US retail sales, according to McKinsey. And customers are ready to delegate: 69% of Americans would let a bank's AI assistant complete at least one financial task for them (TD Bank, 2026).

The same autonomy that makes agents useful makes them dangerous in the wrong hands. Agentic fraud, fraud that AI agents carry out or assist, can run a whole attack loop at machine speed. Agents can generate fake identities and media, submit them, learn from each rejection, and try again, all in seconds. Incode's Agentic Fraud Report already documents 44 independently sourced AI fraud incidents as of H2 2026.

Agentic fraud in this guide means AI-generated, AI-assisted, or agent-operated fraud.

That leaves platforms with a decision they never had to make before. When an agent shows up, do you block it or let it in?

Know Your Agent (KYA) makes a third option possible: admit the agents you can hold accountable, and stop the rest.

How does Know Your Agent (KYA) work?

Know Your Agent (KYA) works as a continuous lifecycle that runs for as long as an agent has access. A complete KYA program identifies the agent, binds it to a verified human or business, scopes what it may do, monitors its behavior and intent, and steps up, revokes, or audits when something changes.

1 Identify the agent. Detect that a session or request comes from an AI agent, then classify it: which agent, which developer or platform, and which surface it uses (web, app, Model Context Protocol (MCP), or agent-to-agent (A2A)).
2 Bind the agent to an accountable principal. Verify the human or business behind the agent and link that verified identity to it. This step turns an anonymous bot into an accountable actor, and binding is only as strong as the verification behind it, so a Know Your Agent (KYA) solution should verify the owner with KYC-grade rigor.
3 Scope what the agent may do. Issue a credential that records the delegated authority: the scope (which actions and accounts), the amount, and the duration. This record is often called the agent mandate. Least privilege, the smallest set of permissions the task needs, is the default.
4 Monitor behavior and intent. Check each action against the agent's declared scope. Flag intent drift, such as a shopping agent that suddenly tries to change a payout account.
5 Step up, revoke, and audit. Before a high-risk action, step-up verification asks the human to verify again, live. Revoke credentials when behavior changes, and keep an audit trail that traces every action back to a verified person.

The Know Your Agent (KYA) lifecycle at a glance

Stage	Question it answers	Typical signals	Risk if skipped
1. Identify	Is this an agent, and which one?	Agent declarations, signed requests, and behavioral and device signals	Agents can pass as humans
2. Bind	Who is accountable for it?	Verified owner identity (liveness, document, government-record match), business verification	The agent's actions have no accountable owner
3. Scope	What may it do?	Scoped token, mandate amount and duration, consent record	The agent can act beyond what the owner approved
4. Monitor	Is it behaving as authorized?	Action-versus-scope checks, velocity, intent drift	Hijacked or rogue agents can go unnoticed
5. Step up, revoke, or audit	Is the human still in control, and can we prove it?	Live re-verification, revocation lists, audit logs	Regulators, disputes, and investigations lack evidence

What three questions does Know Your Agent (KYA) answer?

Every Know Your Agent (KYA) decision comes down to three separate questions, and an agent has to pass all three.

Identity 01

Is this really the agent it claims to be?

Authority 02

Has a verified human or business authorized this agent to take this specific action?

Accountability 03

Can the platform attribute each action to that verified principal and audit it after the fact?

A valid credential can satisfy the first question and fail the other two. That gap is where agent-driven fraud often finds its opening.

How is Know Your Agent (KYA) different from KYC and KYB?

Know Your Agent (KYA) builds on Know Your Customer (KYC) and depends on it. KYC and Know Your Business (KYB) verify the people and companies a platform serves. KYA verifies the software that acts for them and binds it to a verified owner, so the strongest KYA programs start with KYC-grade verification of the human behind the agent.

KYC, KYB, KYA, and KYH compared

Attribute	KYC	KYB	KYA	KYH
Verifies	A person	A business	An AI agent and its accountable owner	Whether a live, real human is present
Core question	Is this person who they claim to be?	Is this company real, and who controls it?	Who is this agent acting for, and what may it do?	Is a real person here right now?
When	Onboarding, periodic review	Onboarding, periodic review	Before an agent gets access, and continuously	At high-risk moments
Typical checks	ID document, selfie, liveness, watchlists	Business registries, beneficial owners	Owner verification, scoped credential, monitoring, audit	Liveness, injection detection, and deepfake detection

Know Your Human (KYH) is the human-side counterpart to KYA: it checks whether a real person is present before an agent acts in that person's name.

How does Know Your Agent (KYA) compare with OAuth, API keys, and agent authentication?

Know Your Agent (KYA) is different from OAuth, API keys, and passkeys. Those credentials confirm that something has valid access; KYA verifies the human owner behind the agent and binds the agent's actions and permissions to that person.

Authentication tools answer one question well: does this caller have valid access? Who stands behind an agent, whether that person approved this specific action, and who answers for it afterward are separate questions.

A stolen API key works the same for an attacker as for its owner. An OAuth grant records that someone consented once; whether a real, verified person gave that consent is a separate check.

Agent protocols leave the same gap. The Model Context Protocol (MCP) standardizes how agents connect to tools, APIs, and data; who the agent is, what it is authorized to do, and who answers when it goes wrong sit outside its scope.

What authentication methods prove

Method	What it proves	What it leaves open
API key	The caller holds a valid key	Who holds it, or whether someone stole it
OAuth token	A user granted scoped access at some point	Whether that user was a real, verified person, and whether they intended this action
Passkey	Possession of a bound device credential	Which agent is acting, and under what limits
MCP authorization	The agent may reach this tool or server	Who the agent is, and who is accountable for it
Know Your Agent (KYA)	A verified human or business is bound to this agent and its permitted actions, with an audit trail	Closes the gaps above when layered on authentication

Is all AI agent traffic fraud?

No. Much of it comes from customers and businesses that sent an agent on purpose: a customer's assistant comparing prices, a business's agent reconciling invoices, or a travel assistant rebooking a missed connection. Treating all of it as hostile can break the experiences customers increasingly expect.

The risk is the agent no one can account for. Agentic fraud depends on anonymity: an attacker's agent can pass as a human, or as someone else's agent, and keep retrying until a check lets it through. Removing that anonymity lets a platform admit agents bound to a verified human, scope what each one may do, and deny the rest.

Know Your Agent (KYA) is the counterpart to agentic fraud. Agent traffic can be legitimate or fraudulent, and KYA lets a platform admit agents acting for a verified, approving human and stops unverified agents from passing as humans.

How Know Your Agent (KYA) decides

When an agent arrives	KYA decision
Bound to a verified owner and acting within scope	Admit and monitor
Bound to a verified owner and attempting a high-risk action	Step up with live re-verification of the owner
Unverified, or claiming to be human	Deny, or route to human verification

Which Know Your Agent (KYA) frameworks and standards exist today?

Agent identity standards are still emerging, and payment networks and standards bodies are both working on them.

Under the approach, each network keeps its own verification and decisioning processes, and each agent is linked to a validated operator, cardholder, or business, assessed against security and behavioral requirements, and monitored continuously (Business Wire).

Each company brings a protocol it has already released: Visa's Trusted Agent Protocol, Mastercard Verifiable Intent, and Ant International's Agentic Mobile Protocol. They will work toward common principles through BuildFin.ai, an industry platform convened by the Monetary Authority of Singapore (MAS), building on its Safeguards for Agentic Finance at Runtime (SAFR) framework.

Know Your Agent (KYA) frameworks and standards

Framework or standard	Who	What it helps verify
KYA interoperability collaboration (exploratory)	Ant International, Mastercard, and Visa (September 2026)	How payment networks identify and onboard AI agents, each linked to a validated operator, cardholder, or business
Trusted Agent Protocol	Visa (October 2025)	Lets merchants tell legitimate AI shopping agents from malicious bots
Verifiable Intent	Mastercard, with Google (March 2026)	A tamper-resistant, cryptographic record of what a person authorized an AI agent to do
Agentic Mobile Protocol	Ant International (April 2026)	Secure, open-source agent payment connections to mobile wallets and apps
HTTP Message Signatures (RFC 9421) and Web Bot Auth	IETF (RFC 9421 in 2024; Web Bot Auth in draft)	That the agent's key signed a request and nobody altered it in transit
Model Context Protocol (MCP)	Agentic AI Foundation at the Linux Foundation (introduced by Anthropic, 2024)	How agents connect to tools and data; agent identity sits outside its scope
Agent2Agent (A2A) protocol	Linux Foundation (introduced by Google, 2025)	How agents discover, communicate, and transact with each other; identity depends on the implementation
Verifiable Credentials 2.0 and Decentralized Identifiers (DIDs)	W3C	Portable, cryptographically checkable claims about an agent or its owner
AI Agent Standards Initiative	NIST CAISI (February 2026)	Interoperable and secure standards for AI agents, still in progress

Standards and frameworks current as of November 2026.

Where is Know Your Agent (KYA) used?

Know Your Agent (KYA) applies wherever an AI agent acts for someone else, and three settings show why it matters.

Agentic payments

An agent that initiates a transaction should carry proof of who authorized it and under what mandate. With that proof, a merchant can tell a customer's shopping agent from a card-testing bot.

Customer-facing agents

Agents that log in, change settings, or file claims on a customer's account need the same assurance as the customer, plus a record of what they were allowed to do. See how platforms verify AI agents acting for customers.

Regulated industries

In financial services, healthcare, and legal workflows, every action needs an accountable party. KYA keeps the audit trail that shows who authorized each action when a regulator asks.

Across all three, KYA applies at the same control points:

Agent onboarding. Verify the human owner once, before the agent gets any credential.
Sensitive-action step-up. Re-verify the person live when an agent goes over its limit or changes account access.
Multi-agent oversight. See every agent one verified person runs.
Incident response and audit. Trace any agent action back to a real, verified person.

How should you evaluate a Know Your Agent (KYA) approach?

Judge any Know Your Agent (KYA) approach on seven criteria, and for each one, ask what ships today and what is still on the roadmap.

1 Human-owner verification strength. Does it verify the owner with liveness, deepfake detection, and injection-attack detection, or only with an email or OAuth login?
2 Agent detection across surfaces. Does it recognize agents on web, app, MCP, and agent-to-agent traffic?
3 Bound, scoped credentials. Is the credential cryptographically tied to the verified owner and to specific actions, amounts, and durations?
4 Step-up at high-risk actions. Can it require live re-verification before a payment over the limit or an access change?
5 Continuous monitoring. Does it watch for intent drift and hijacked agents after onboarding?
6 Revocation and audit. Can you revoke an agent instantly and produce an audit trail you can hand to a regulator or a court?
7 Privacy and data minimization. What personal data does it store, where, and for how long?

How does Incode Know Your Agent verify AI agents?

Incode Know Your Agent ties every AI agent to a verified human and the mandate they set, and keeps every action traceable to that person. It is the end-to-end trust layer for AI agents, built on human identity, and it runs through five components that give a business access, control, and visibility over each agent.

1Agent Detection finds the AI agents in your traffic based on multiple device, behavioral, and network signals.
2Agentic Identity ties each agent to a live, verified human and the mandate they set: scope, amount, and duration.
3Agent Access Management checks every action against the verified mandate, and anything over the limit waits for the person's live approval.
4Agent Monitoring blocks compromised or misaligned agents across sessions and organizations.
5Agent Audit keeps every action traceable to the person who authorized it, under the policy the business sets.

The person verifies once with a live face and gives consent.

Deepsight checks whether a real person is present, with deepfake and injection-attack detection.
GovFaceMatch matches the face to the government's own record.
TrustGraph reads the person's history across the network.

In Purdue University testing, Deepsight had 0.2% false acceptance on deepfakes, compared with 13% for the next vendor. More than 800 organizations trust Incode, including nine of the top 10 US banks, eight of the top nine US telcos, and four of the top five global marketplaces, and TrustGraph links people, agents, and actions across them. Gartner named Incode a Leader in the Gartner® Magic Quadrant™ for Identity Verification for the third consecutive year.

Request a demo

FAQ

Frequently asked questions about Know Your Agent (KYA)

Short answers to the questions teams ask most when they start planning how to verify AI agents.

What is Know Your Agent (KYA)?

Know Your Agent (KYA) is the practice of verifying an AI agent and binding it to an accountable human or business before it is allowed to act. KYA answers three questions: who the agent is, which verified human or business stands behind it, and what it may do.

What does KYA stand for in AI?

In AI, identity, and payments, KYA stands for Know Your Agent: the verification of AI agents and the people or businesses they act for. Real estate and insurance use the same phrase for vetting human agents, a separate practice.

How is KYA different from KYC?

Know Your Customer (KYC) verifies a person. Know Your Agent (KYA) verifies an AI agent and binds it to the person or business accountable for it. KYA builds on KYC: the platform verifies the agent's owner first, and the agent then carries a scoped, auditable link to that verified identity.

How do platforms verify AI agents?

Platforms verify AI agents in layers. First they identify the agent and bind it to a verified human or business owner. Then they issue a scoped credential, monitor the agent's behavior against that scope, and require step-up verification or revoke access the moment something changes. Every action goes into an audit trail that traces it back to the verified owner.

Why aren't OAuth, API keys, or passkeys enough to verify an AI agent?

OAuth, API keys, and passkeys confirm that an agent has valid access, and they leave open who stands behind it. A stolen key works for an attacker exactly as it works for its owner. Know Your Agent (KYA) verifies the human owner behind the agent and binds its actions and permissions to that person.

Do AI agents need KYC?

An AI agent has no legal identity, so it cannot complete KYC itself. The platform verifies the human or business that owns the agent instead, then binds the agent to that verified identity with scoped permissions. That binding is the core of Know Your Agent (KYA).

What is agentic identity, and how does it relate to KYA?

Agentic identity is the verified digital identity of an autonomous AI agent: confirming who operates it, what it's authorized to do, and ensuring its actions are traceable and accountable across enterprise systems. Know Your Agent (KYA) is the framework for establishing that identity and keeping it trustworthy for as long as the agent acts.

Is all AI agent traffic fraudulent?

No. Much AI agent traffic comes from real customers and businesses. Know Your Agent (KYA) lets a platform admit agents acting for a verified, approving human and stops unverified agents from passing as humans, the tactic agentic fraud depends on.

What is the Visa, Mastercard, and Ant International KYA framework?

On September 9, 2026, Ant International, Mastercard, and Visa announced a collaboration on Know Your Agent (KYA) interoperability to help card networks, digital wallets, agent platforms, and marketplaces identify and onboard AI agents across networks. Each brings its own protocol: Visa's Trusted Agent Protocol, Mastercard Verifiable Intent, and Ant International's Agentic Mobile Protocol. The work runs through BuildFin.ai, convened by the Monetary Authority of Singapore, and is still exploratory.

Who is accountable when an AI agent acts?

Under a Know Your Agent (KYA) model, the verified human or business the agent is bound to is the accountable party. KYA creates the audit trail that links each action to that principal, so platforms can resolve disputes and answer regulators.

Does MCP verify who an AI agent is?

No. The Model Context Protocol (MCP) standardizes how agents connect to tools, APIs, and data; who an agent is, what it is authorized to do, and who answers when it goes wrong sit outside its scope. Know Your Agent (KYA) adds verified owner identity on top of MCP connections.

How does KYA help stop agentic fraud?

Agentic fraud relies on anonymous agents that pass as humans and retry until a check lets them through. Know Your Agent (KYA) removes that anonymity: it requires a verified, accountable owner for every admitted agent, checks each action against that owner's mandate, and asks for live re-verification before high-risk actions.

Go deeper on agentic fraud and Know Your Agent (KYA)

Ready to see how Incode Know Your Agent ties every AI agent to a verified human?

Request a demo

This guide is for general information and is not legal or compliance advice. Requirements for verifying AI agents vary by jurisdiction and industry.