Can cryptographic signatures help merchants identify trusted AI agents?

Visa describes Trusted Agent Protocol, developed with Cloudflare, as "a foundational framework for agentic commerce that enables secure communication between AI agents and merchants during every step of a transaction." The press release frames the problem as a recognition gap: as AI agents browse and buy for consumers, merchants need, in Visa's words, "a way to recognize trusted agents, verify their credentials, and maintain customer relationships without overhauling their systems," and to tell commerce-intent agents apart from "malicious automation and rogue bots." Jack Forestell, Visa's Chief Product & Strategy Officer, says the protocol is "focused on creating no-code functionality for merchants to securely identify agents with an intent to buy and provide a better payments and personalized experience for its known users."

The mechanism Visa describes is agent-specific cryptographic signatures carrying three data elements: Agent Intent ("an indication that the agent is a trusted agent with an intent to retrieve additional details about, or purchase, a specific product"), Consumer Recognition (whether the consumer has an existing account or prior interaction with the merchant), and Payment Information (optional payment data to support the merchant's preferred checkout method). Visa says the protocol is built upon the foundational HTTP Message Signature standard, and aligned with Web Bot Auth, which it presents as requiring "minimal user experience (UX) changes" on merchant sites. Visa states it is "committed to aligning closely with global standards bodies like IETF, OpenID Foundation and EMVCo," and names the Agentic Commerce Protocol and Coinbase's x402 as protocols it intends to complement rather than replace.

Trusted Agent Protocol sits alongside, but is not the same mechanism as, the other agent-trust proposals in the library: where How can AI agents prove who authorized their actions? anchors trust in tokens that trace an agent's authority back to a specific human principal, Visa's protocol anchors trust in a merchant-facing signature that asserts intent and known-customer status without describing a chain back to a person. It also differs in audience from Can a protocol split agent access from agent incentives?, which governs what a consumer's own agent may do across channels; Visa's protocol instead governs how a merchant recognizes an incoming agent as legitimate. Visa's own Does naming a payments brand increase trust in AI transactions? supplies a motive for TAP: a trust gap between AI-assistant adoption and willingness to let AI pay, which a merchant-recognized, brand-anchored protocol is positioned to narrow. By contrast, Can existing identity standards work for AI agents? argues for reusing OAuth, OIDC and SPIFFE/SPIRE rather than creating agent-specific mechanisms — which is exactly the kind of new, purpose-built protocol NIST's concept paper would caution against, even as Visa frames TAP as compatible with existing web infrastructure.

The excerpt is a vendor press release, not an independent evaluation: Visa's claims about Cloudflare's role, partner feedback, and standards-body alignment are company statements, not verified outcomes, and the release does not report real-world adoption, transaction volumes, or dispute/fraud data from using the protocol. It also cites, without attributing a source, that "AI-driven traffic to retail websites in the United States surged over 4,700%" and that "85% of shoppers who have used AI to shop say it improved their shopping experience" — figures that belong to Visa's framing of the problem, not to a named study. The release states explicitly that "these initial specifications apply to the Visa network in this phase," so interoperability with the Agentic Commerce Protocol and x402 is a stated intention rather than a demonstrated integration.