Visa Trusted Agent Protocol (TAP): What It Is and How It Works

The Visa Trusted Agent Protocol (TAP) is Visa's answer to a foundational agentic commerce question: how does the payment network know that the agent attempting a transaction is who it claims to be, acting with real consumer authorization?

TAP is a Visa-network protocol that does two things at once: (1) verifies the AI agent's identity ("who is the agent") and (2) verifies the consumer's authorization ("who is the consumer"). Both checks happen at transaction time, before the payment is authorized.

By early 2026, Visa TAP had attracted 100+ partners, with 30+ partners actively in sandbox testing and 20+ integrating in production. Commercial launch was set for Q1 2026 (Visa public statements, late 2025 / early 2026). Akamai partnered with Visa in December 2025 to bring TAP into the edge security layer, allowing dual-identity verification at the CDN level (Visa-Akamai partnership announcement, Dec 17 2025).

Fiserv was the first major payments processor to adopt TAP at scale (announced January 2 2026), wiring TAP into its tokenization, authentication, and fraud-prevention pipelines so any merchant on Fiserv processing inherits TAP-compliant agent payments without separate integration.

TAP layers two new checks on top of the existing Visa authorization flow:

Agent identity verification. The AI agent (e.g. ChatGPT, Google AI Mode, a custom enterprise agent) presents a verifiable identity credential to the network. Visa's registry checks that the agent is enrolled in the TAP program, that its credential is current, and that it has not been revoked.

Consumer authorization verification. Separately, the network checks that the consumer signed an authorization for this specific transaction scope. Visa's implementation supports Mandate-style cryptographic authorization (the same primitive used in AP2) and Visa's native tokenization framework.

Fraud rules tuned for agents. TAP transactions get scored against agent-specific fraud rules: velocity patterns, mandate-vs-transaction matching, anomaly detection on agent behavior. This is materially different from the human-shopper fraud rule set Visa has run for decades.

From the merchant's perspective, TAP is invisible at integration time when the merchant's processor (e.g. Fiserv, Worldpay) supports it. The protocol runs at the network layer; the merchant submits a normal authorization request and gets back an approve / decline with TAP signals included in the response.

The three payment-layer protocols emerging in 2025-2026 differ in scope and ownership:

Most retailers will touch all three indirectly through their processor. Visa transactions clear via TAP, Mastercard via Agent Pay, and AP2 sits above both as an authorization layer processors can use across networks.

Retailers receive Visa TAP through their processor. The integration work happens at the processor and network layer. What matters for the retailer is what TAP enables:

1. Agent transactions clear without manual review. TAP-verified transactions get the same authorization-time scoring as human-initiated card transactions, so agent purchases flow through the same review as any other.
2. Chargeback defense. Mandate-based authorization is auditable. Disputes over "the agent did this without my permission" now have a clear cryptographic record.
3. Higher transaction limits. Without strong agent verification, networks and processors keep agent transactions small. TAP allows higher-value agent purchases (recurring orders, subscriptions, restocking) with the same safeguards.

The practical move for a retailer is to confirm that your processor supports TAP (Fiserv was the first major one, others are following through 2026) and to leave the rest to the payment stack. The part of agent commerce that does need your attention is upstream of payment: whether the agent finds and understands your products in the first place. That depends on the product information behind each listing, which is the work Paz handles for ecommerce teams.