Mastercard's Agent Pay is the payments giant's full-stack framework for AI-initiated transactions, combining agentic tokens, a merchant acceptance framework, and the open-sourced Verifiable Intent standard into a single trust layer for agentic commerce. In a market where every major player is racing to define how AI agents spend money, Mastercard is betting that the network closest to merchants and issuers wins.
Founded 1966 | HQ: Purchase, New York | Public (NYSE: MA) | $31.5 billion annual revenue
MM Verified
Overview
Mastercard launched Agent Pay in April 2025 as its agentic payments programme, introducing Mastercard Agentic Tokens: dynamic digital credentials that allow AI agents to transact on a consumer's behalf using the same tokenisation technology that secures mobile payments globally. By March 2026, Agent Pay has completed live transactions across four continents.
The programme has three layers. Agentic Tokens handle delegated payment credentials. The Agent Pay Acceptance Framework establishes standards for how merchants verify and accept agent-initiated transactions. And Verifiable Intent, open-sourced in March 2026, creates cryptographic proof that an agent acted within a consumer's explicit instructions.
US issuers including Citi and US Bank were enabled in late 2025, with all US Mastercard cardholders equipped by November 2025. Live pilots have completed in Australia (with Commonwealth Bank and Westpac), Malaysia (with CIMB, Maybank, and RHB), and Europe (with Santander). OpenAI integrated Agent Pay into ChatGPT's Instant Checkout, and PayPal announced wallet integration for agent-initiated transactions.
What We Like
Network-level trust infrastructure. Mastercard is not building a standalone product. Agent Pay extends the same tokenisation and settlement infrastructure that processes billions of transactions annually. Agentic Tokens inherit the security properties of network tokens, meaning merchants and issuers do not need new systems to support them.
Verifiable Intent fills the accountability gap. As we explored in our analysis of Mastercard's Verifiable Intent framework, no other protocol answers the question: did the agent do what the human asked? Built on FIDO Alliance, EMVCo, and W3C standards, open-sourcing it on GitHub signals Mastercard wants this as shared infrastructure, not a proprietary moat.
Protocol interoperability is deliberate. Agent Pay works with Google's Universal Commerce Protocol, OpenAI's Agentic Commerce Protocol, and the Agent Payments Protocol (AP2). Google's VP of Payments described Verifiable Intent as "a natural accelerator for scaling agentic commerce," according to PYMNTS.
Fiserv integration reaches the merchant long tail. As we covered in our analysis of the Fiserv-Mastercard partnership, Fiserv processes more US merchant transactions than any other acquirer. Agent Pay in Fiserv's Clover platform means millions of merchants can accept agent-initiated transactions without custom development.
4,000 global advisors support adoption. Mastercard's Agent Suite, launched in January 2026, combines customisable AI agents with consulting services backed by 4,000 global advisors.
What to Watch
Not self-serve. Unlike Stripe's Agent Toolkit, which installs via npm, Mastercard Agent Pay requires issuer enablement, acquirer integration, and network-level provisioning. Developers cannot test it in a sandbox today and ship tomorrow.
Documentation lags behind the product. Mastercard's developer portal covers Agent Pay at a high level, but the depth does not match what Stripe publishes at docs.stripe.com. Detailed implementation guides and working code samples are essential for a product this new.
Pricing is opaque. Mastercard has not disclosed specific pricing for Agent Pay services. The incremental cost of agentic tokens, Agent Suite consulting, and acceptance framework integration is not publicly documented.
Dispute resolution is untested at scale. Verifiable Intent creates the records needed to adjudicate agentic commerce disputes, but no high-volume dispute cycle has stress-tested them in production.
Pricing & Deployment
Mastercard Agent Pay pricing is not publicly disclosed. Transactions flow through standard Mastercard interchange and network fee structures via existing acquirer relationships. Agent Suite consulting and agentic token provisioning are available through direct engagement with Mastercard. Deployment requires issuer enablement and acquirer integration, with Fiserv, Braintree, and Checkout.com among the first acquirers embedding Agent Pay token support into their existing tokenisation services.
Compliance & Security
Mastercard Agent Pay inherits the full compliance posture of the Mastercard network: PCI DSS compliance, EMVCo tokenisation standards, and global regulatory coverage across 210+ countries and territories. Verifiable Intent is built on FIDO Alliance, W3C, and IETF standards. Selective Disclosure ensures no single party in the transaction chain sees the complete consumer profile. All agentic tokens are network-issued, time-scoped, and revocable.
Verdict
Mastercard Agent Pay is the most comprehensive network-level agentic commerce framework available today. For issuers, acquirers, and large merchants within the card network ecosystem, it offers proven tokenisation, a merchant acceptance standard, and the only open-sourced intent verification layer in the market.
This is not the right choice for startups wanting to prototype agent commerce in a weekend. Stripe's Agent Toolkit offers faster time-to-value for developer-led implementations. But for enterprises evaluating infrastructure that scales with the card networks, Agent Pay is the platform to watch.
With live transactions across four continents and Verifiable Intent positioning as the shared trust standard, Mastercard is building the layer between consumer intent and machine action.
Explore Mastercard Agent Pay: mastercard.com/agent-pay
How we scored it
| Criterion | Score | Notes |
|---|---|---|
Accuracy & Effectiveness 20% weight | 4.5 | Live transactions on four continents; proven tokenisation infrastructure |
Compliance & Security 15% weight | 5.0 | PCI DSS, EMVCo, FIDO Alliance, W3C standards; Selective Disclosure |
Documentation 15% weight | 3.5 | Developer portal exists but lacks depth; Verifiable Intent going open source |
Ease of Setup 10% weight | 3.5 | Requires issuer and acquirer enablement; not self-serve |
Integration Flexibility 10% weight | 5.0 | Works with Google UCP, OpenAI ACP, AP2, Fiserv, PayPal |
Support Quality 10% weight | 4.5 | 4,000 global advisors; Agent Suite consulting from Q2 2026 |
Scalability 10% weight | 5.0 | Global card network processing billions of transactions annually |
Pricing Transparency 10% weight | 2.0 | No public pricing; enterprise-only through acquirer relationships |
Pros
- Network-level trust infrastructure
- Verifiable Intent fills the accountability gap
- Protocol interoperability is deliberate
- Fiserv integration reaches the merchant long tail
- 4,000 global advisors support adoption
Cons
- Not self-serve
- Documentation lags behind the product
- Pricing is opaque
- Dispute resolution is untested at scale
Sources
- Mastercard: Agent Pay Launch Press Release (April 2025)
- PYMNTS: Mastercard Unveils Open Standard to Verify AI Agent Transactions
- PYMNTS: Fiserv-Mastercard Expand Partnership for AI-Initiated Commerce
- Mastercard: Agent Suite Launch (January 2026)
- Mastercard: Australia's First Authenticated Agentic Transactions
- Mastercard: Santander and Mastercard Complete Europe's First AI-Executed Payment
- Google Developers Blog: Under the Hood: Universal Commerce Protocol
- Mastercard: Building Trust in AI Commerce: Agentic Protocols
If AI agents are going to spend trillions on our behalf, the trust infrastructure cannot be an afterthought. Mastercard is building that layer. The question is whether the rest of the industry treats it as shared infrastructure or builds competing alternatives.
Charlie Major is a Product Development Manager at Mastercard. The views and opinions expressed in Major Matters are his own and do not represent those of Mastercard.