Same Job, Different Networks
Visa TAP (Trusted Agent Protocol) and Mastercard Agent Pay are the two card-network protocols for AI-agent-mediated payments. They share the high-level goal: let an AI agent transact on a cardholder's behalf within issuer-defined and user-defined constraints, with the resulting transaction carrying agent-context metadata that downstream parties can recognise. They differ in network ownership, issuer participation, specific metadata schemas, and pilot composition.
At a Glance
| Dimension | Visa TAP | Mastercard Agent Pay |
|---|---|---|
| Network | Visa | Mastercard |
| Status (April 2026) | Pilot, broader rollout planned 2026 | Pilot, broader rollout planned 2026 |
| Pilot issuer composition | Chase, Capital One, several US and international banks | HSBC, Standard Chartered, several European and Latin American banks |
| Geographic strength | Strong in US, balanced internationally | Strong in Europe, Latin America, parts of Asia |
| Agent platform support | Major agent platforms participating | Same major agent platforms |
| Settlement rails | Visa card network | Mastercard card network |
| Merchant cost | Standard Visa rates | Standard Mastercard rates |
Functional Similarity
Functionally, the two protocols are similar enough that brand and merchant teams typically handle them through a single agent-readiness implementation. Both protocols issue agent credentials tied to a cardholder, both evaluate transactions against issuer-defined and user-defined constraints, both produce transactions carrying agent-context metadata. The metadata schemas differ slightly but the functional information is parallel.
The two networks have signalled intent to converge on metadata schemas where possible to reduce merchant integration complexity. As of April 2026 the convergence is partial; merchants accepting both networks should expect to handle minor protocol-specific differences for the next 12-18 months.
Where Visa TAP Has the Edge
Visa's issuer base is larger in the US, which means TAP has more pilot reach in the largest single agent commerce market. The pilot includes Chase and Capital One, two of the largest US issuers, which gives TAP-credentialed traffic disproportionate volume in US-centric agent flows. Brands and merchants serving primarily US consumers should optimise for TAP first.
Where Mastercard Agent Pay Has the Edge
Mastercard's issuer base is stronger outside the US, particularly in Europe, Latin America, and parts of Asia. Agent Pay's pilot includes HSBC and Standard Chartered, both globally significant banks with strong international issuer footprints. Brands and merchants with meaningful non-US agent traffic should not optimise only for Visa TAP; Agent Pay's reach in international markets is the differentiating factor.
Should Merchants Treat Them Differently?
For most merchants, no. The handling is similar enough that a single agent-readiness implementation covers both networks with minor protocol-specific metadata differences. The agentic commerce readiness scorecard (six dimensions: render-without-auth, schema, price/stock consistency, anti-bot infrastructure, order review brand identity, payment authorization) covers what works for both.
The exception is brands operating in markets where one network dominates significantly. In Brazil where Mastercard is dominant, Agent Pay-specific optimisation pays back faster than TAP-specific optimisation. In the US where Visa is roughly co-equal with Mastercard, both protocols matter roughly equally.
How They Compose with Other Protocols
TAP and Agent Pay sit at the card-network rail level. They are accessed through higher-layer protocols (AP2 most prominently) that mediate agent intent and authorisation in a network-agnostic way. From an agent-platform perspective, transacting through AP2 may settle through TAP for Visa-credentialed users and through Agent Pay for Mastercard-credentialed users; the agent layer does not need to handle the network distinction directly.
How Presenc AI Tracks Both
Presenc AI tracks brand visibility in agent-mediated commerce flows across TAP, Agent Pay, x402, MPP, and AP2 with separate scoring per protocol. The decomposed view lets brands see which protocol-specific surfaces are working and which are not. For most brands, the protocol breakdown is less important than the overall agentic commerce readiness; the breakdown matters mainly for international brands where geographic mix interacts with network mix.