Trusted Agent Protocol
The Trusted Agent Protocol is an approach by Visa in which AI agents identify themselves to merchants with signed requests and pass on details about the customer and payment.
Visa introduced the Trusted Agent Protocol in October 2025, developed together with Cloudflare. It builds on Web Bot Auth and HTTP Message Signatures as defined in RFC 9421. The signatures are bound to merchant and purpose and limited in time, so they cannot be replayed or relayed.
Besides the agent's identity, the protocol can carry indications about the customer and the payment data. Visa provides the specification and sample code on GitHub; use is subject to its own product terms.
Example
A shopping agent opens a merchant's product page and signs each request for exactly this merchant. The merchant verifies the signature with the key from Visa's directory, recognizes an approved agent and offers checkout instead of blocking the request as a bot.
How it differs
The Trusted Agent Protocol focuses on the agent toward the merchant. Verifiable Intent and AP2, by contrast, prove the user's intent and limits for the purchase. In practice, the two levels complement each other.
In practice
- Review Visa's product terms before using code from the repository.
- Treat the signature as the agent's identification and check the purchase limits additionally.
- Keep signed requests and verification results, so disputes with customers can be backed by evidence.
Role in KYA
KYA verifies the underlying HTTP signature following RFC 9421. A dedicated profile for the protocol's additional fields is not yet implemented, because the sample code is subject to its own product terms.
Related check: Issuer (issuer_resolution). Is the key bound to the issuer – through a DID or because the verifying party trusts it?
Frequently asked questions
What is the Trusted Agent Protocol?
An approach by Visa in which AI agents identify themselves to merchants. It uses signed HTTP requests following RFC 9421 and Web Bot Auth, bound to merchant, purpose and time window.
How does it differ from Verifiable Intent?
The Trusted Agent Protocol mainly identifies the agent toward the merchant. Verifiable Intent proves the user's signed purchase intent with its limits. Both levels can appear together.
Can KYA verify the Trusted Agent Protocol?
The underlying HTTP signature following RFC 9421, yes. KYA does not verify the protocol's additional fields in this version.