Know Your Agent (KYA): 6 Startups Verifying AI Agents Before They Pay

A passport control booth in an empty terminal at night, one booth lit warm

Know Your Agent (KYA) is the check a merchant or bank runs on an AI agent before it lets the agent buy: who built it, who it acts for and what it may spend. How KYA differs from KYC, the card-network programs, and six startups building it.

3 questions which agent, acting for whom, allowed to do what
$8.3M largest seed here: Nava, April 2026
6 startups compared
2026-11-22 facts last verified

Key Takeaways

  • KYA answers three questions KYC never asked. — Which software is this, which verified person or company is it acting for, and what has that principal allowed it to do. KYC stops at the first person; KYA follows the delegation chain.
  • The card networks set the floor. — Visa’s Trusted Agent Protocol and Mastercard’s Agent Pay give registered agents a way to identify themselves to merchants. Startups sell what sits around that: intent checks, limits, escrow and logs.
  • The category is early and small. — The six startups here have raised seed and pre-seed rounds in 2026, between $2.5M and $8.3M each. Expect most merchants to start with their fraud vendor’s bot controls before buying a dedicated KYA product.

Know Your Agent (KYA) is the check that comes before an AI agent is allowed to pay. A merchant, bank or platform asks which agent this is, which verified person or company it is acting for, and what it has been allowed to do. This page explains the idea and compares six startups building it.

What Know Your Agent means

Know Your Customer (KYC) verifies a person. Once that person hands a task to an agent, the merchant sees software, not a person, and the old checks break: there is no face to match and no one-time code an agent can read from a phone. KYA fills the gap with three links. The agent proves what it is with a signed credential. That credential points to a verified principal. And the principal’s permissions, such as a spending cap or a list of allowed merchants, travel with the request so the merchant can check the purchase against them.

What Visa and Mastercard already do

The card networks set the base layer. Visa’s Trusted Agent Protocol, launched in October 2025 with Cloudflare, lets approved agents sign their requests so merchants can tell them apart from bad bots. Mastercard’s Agent Pay registers agents and issues agentic tokens that bind a card credential to one agent. Google’s Agent Payments Protocol (AP2) carries a signed mandate that proves the user approved the spend; see ACP vs AP2. The startups below build on top: intent checks, limits, escrow and audit trails.

The lead list comes from Alan Lin’s November 2026 map of agentic commerce startups, which counts $23M raised in 2026 by identity and checks companies. We checked each company and round below against the announcement.

The six at a glance

CompanyWhat it checksLatest disclosed fundingBest fit
NavaEscrow and intent checks$8.3M seed, April 2026; co-led by Polychain and ArchetypeFinancial agents that move money on a user’s instruction.
BelticAgent traffic verification$7.3M seed, September 2026; led by Norwest (about $8.8M total)Merchants and platforms seeing agent traffic at checkout.
t54 LabsIdentity, risk and compliance$5M seed, February 2026; co-led by Anagram, PL Capital and Franklin Templeton, with RippleCrypto and fintech platforms opening accounts to agents.
RalioBusiness payment controls$2.5M pre-seed, April 2026; led by Sure Valley VenturesFinance teams letting agents pay suppliers.
SkyfireKYA tokens plus payments$9.5M total, 2024; Coinbase Ventures and a16z CSXSellers that want verified agents to pay for content or APIs.
PayOSAgentic tokens at checkoutNot disclosedMerchants preparing card checkout for registered agents.

The startups

Nava holds an agent’s funds in a blockchain-based escrow and releases them only after the transaction is checked against what the user asked for. Funding: $8.3M seed, April 2026; co-led by Polychain and Archetype (source). Best fit: Financial agents that move money on a user’s instruction.

Beltic

Beltic verifies AI agents that visit a business and checks each action against the business’s rules, then allows it, blocks it or asks for confirmation. Funding: $7.3M seed, September 2026; led by Norwest (about $8.8M total) (source). Best fit: Merchants and platforms seeing agent traffic at checkout. Website.

t54 Labs

t54 Labs builds identity, risk and compliance tools for autonomous agents that transact in financial markets and payments. Funding: $5M seed, February 2026; co-led by Anagram, PL Capital and Franklin Templeton, with Ripple (source). Best fit: Crypto and fintech platforms opening accounts to agents. Website.

Ralio

UK-based Ralio sits between AI agents and payment rails for business payments, verifying intent, enforcing limits and logging each decision. Funding: $2.5M pre-seed, April 2026; led by Sure Valley Ventures (source). Best fit: Finance teams letting agents pay suppliers. Website.

Skyfire

Skyfire issues Know Your Agent identity tokens that an agent presents to a seller, alongside its payment network for agents. Funding: $9.5M total, 2024; Coinbase Ventures and a16z CSX (source). Best fit: Sellers that want verified agents to pay for content or APIs. Website.

PayOS

PayOS completed one of the first live agent payments on Mastercard’s network using a Mastercard Agentic Token, which binds a card credential to a registered agent. Funding: Not disclosed (source). Best fit: Merchants preparing card checkout for registered agents.

A KYA checklist for merchants

  1. Identify the agent. Accept signed agent credentials (Visa Trusted Agent Protocol, Mastercard agentic tokens, or a verified-bot list) and treat unsigned automation as a bot.
  2. Link it to a verified principal. Every agent purchase should trace to a customer you have already verified.
  3. Check the mandate. Compare amount, merchant and item against the permissions the principal granted.
  4. Escalate the edge cases. Ask the person to confirm anything outside the mandate rather than declining outright.
  5. Log it. Keep the instruction, mandate, agent identity and outcome for disputes. For the payment side, see agentic payments.

Sources

  • Funding announcements linked in each entry above, read 2026-11-22.
  • Lead list and segment total: Alan Lin, “85 startups building to advance agentic commerce” (startup map, November 2026; his sources: company sites, Dealroom and funding announcements to November 20, 2026).
  • US search volumes from Google Ads data via DataForSEO, November 2026.

What does Know Your Agent (KYA) mean?

Know Your Agent is the process of verifying an AI agent before letting it transact: confirming which software it is, which verified person or business it acts for, and what that principal has authorized it to do. It extends Know Your Customer (KYC) from people to the software acting for them.

How is KYA different from KYC?

KYC verifies a person or company once, at onboarding. KYA verifies a piece of software each time it acts, and links it back to a KYC-verified principal plus a set of permissions such as a spending limit, allowed merchants or an expiry. A KYC-verified customer can run many agents, each with its own limits.

How do merchants verify AI agents today?

Mostly through card-network programs and bot management. Visa’s Trusted Agent Protocol and Mastercard’s Agent Pay let registered agents identify themselves with signed credentials or agentic tokens, and bot-management providers such as Cloudflare list verified shopping agents. Dedicated KYA startups add intent checks, limits and audit logs.

Who is liable if an AI agent makes a wrong purchase?

It is not settled. Card-network agent programs tie a purchase to a registered agent and a user’s mandate, which helps assign responsibility in a dispute, but rules on chargebacks and liability for agent-initiated payments are still being written. Keep a record of the instruction, the mandate and the agent’s identity for every purchase.

Explore More

Ready to Find the Right AI Tools?

Browse our data-driven rankings to find the best AI tools for your team.