Mastercard has moved aggressively to make agentic commerce, transactions carried out by AI agents on behalf of people, safer, more interoperable and easier to adopt across the payments ecosystem. Its Agent Pay program, announced in 2025, introduces agentic tokens, new acceptance frameworks and developer tooling aimed at embedding cryptographic consent, traceability and issuer visibility into AI-driven purchases.
Those moves come amid a broader industry push toward common protocols for agent-driven shopping: Google unveiled the Universal Commerce Protocol in January 2026 and a growing set of retailers and payments networks have endorsed open standards to reduce bespoke integrations and preserve merchant control. The result is an emerging standards ecosystem where networks, platforms and device makers can interoperate so AI agents can discover, negotiate and pay on behalf of users.
What is agentic commerce?
Agentic commerce describes a model where software agents, personal assistants, task-specific bots or autonomous services, act on behalf of a human to search, compare, book and purchase goods and services. These agents can live in chatbots, voice assistants, or background orchestration layers and may interact with multiple merchants and services without a human re-entering credentials at checkout.
The model promises convenience: agents can weigh preferences, budgets, timing and seller rules to make optimized choices, and can even manage post-purchase tasks such as returns or subscription adjustments. But that convenience raises new questions about authorization, auditing and the chain of responsibility when purchases occur without a direct, human touch.
To scale safely, agentic commerce requires technical and governance building blocks, tokenized credentials tied to explicit permissions, interoperable checkout messages, and standards for agent identity, logging and dispute resolution. Without them, merchants, issuers and regulators risk fragmentation and unpredictable fraud exposure.
Mastercard’s Agent Pay and agentic tokens
Mastercard launched its Agent Pay program in 2025 to embed payments into agentic experiences and to extend the company’s tokenization work into AI-driven flows. Agent Pay introduces Agentic Tokens, dynamic, permissioned credentials that allow an AI agent to transact within limits set by the cardholder while preserving issuer visibility and fraud protections.
These tokens build on existing tokenization and passkey technology to ensure each agent-initiated payment carries cryptographic proof of intent and adheres to merchant and issuer policies. Mastercard positions these tokens as a way to keep the cardholder in control while enabling seamless delegated payment in conversational and ambient commerce surfaces.
In practice, Agent Pay is designed to let issuers, wallets and platforms recognize an approved agent and the permissions granted to it, so transactions can be authorized, routed and disputed with the same traceability expectations used for traditional card-on-file or contactless payments.
Developer tools and protocol support
To accelerate adoption, Mastercard introduced developer-focused tooling such as an Agent Toolkit and machine-readable API documentation surfaced via a Model Context Protocol (MCP) server. Those tools are intended to make Mastercard APIs discoverable and integrable by AI agents and automation platforms, lowering engineering friction for agentic workflows.
Mastercard’s approach explicitly complements existing and emerging protocols, Agent2Agent messaging, AP2-style payment handoffs and MCP-style context sharing, so agents can understand product data, pricing and order state without fragile point-to-point integrations. The toolkit and Agent Sign-Up onboarding aim to standardize how agents identify themselves to networks and merchants.
For developers and platforms, this means reference implementations and structured documentation will be available to help onboard agents and enable delegated payment flows while preserving security checks that issuers and acquirers require.
Industry collaborations and partnerships
Mastercard has worked with major tech and payments players to ensure its agentic stack fits into larger standards work. Early partnerships announced by Mastercard include collaborations with Microsoft and other AI platform vendors to surface Agent Pay in conversational checkouts and B2B orchestration.
In late 2025, Mastercard and PayPal expanded a partnership to integrate Agent Pay into PayPal’s wallet, demonstrating how wallets and networks can be combined to scale agentic transactions across hundreds of millions of consumers and merchants. That deal underscores the industry desire to align wallets, networks and platforms behind interoperable delegated-payment mechanics.
At the same time, Mastercard’s public endorsements and participation in industry protocols, alongside players such as Google, major retailers and other networks, signal a coordinated push to avoid isolated, incompatible agentic silos and to preserve merchant control of checkout logic.
Regional rollouts and pilot programs
Mastercard has moved deliberately from proofs of concept to pilots in specific markets. The company showcased regional rollouts, such as Agent Pay introductions in Latin America and the Caribbean, and committed to enabling issuers and acquirers in those regions to support agentic transactions under an Agent Pay Acceptance Framework. Those regional efforts illustrate how network-level coordination can accelerate real-world uptake.
Outside Latin America, pilots and commercial deployments have appeared in other markets too, with partners ranging from major retailers to fintechs and acquirers. The cross-market strategy helps test regulatory, merchant and consumer behaviors under different rules and infrastructural constraints.
These pilots also reveal practical integration work, mapping local payment rails, settling tokenization approaches, and aligning issuer risk models, that must be accomplished before broad consumer-facing rollouts become seamless experiences across borders.
Standards landscape and Google’s Universal Commerce Protocol
Google’s Universal Commerce Protocol (UCP), introduced in January 2026, defines a common language for discovery, checkout, payments and order management across agentic surfaces; the protocol has the backing of more than 20 partners, including major retailers and payments firms. UCP is designed to interoperate with agentic payments protocols and to reduce N×N point-to-point integrations. Mastercard’s work, Agent Pay and related tooling, aligns with this multi-vendor standardization trend.
Compatibility between network-level features (like Agentic Tokens) and open commerce protocols is critical: merchants want to remain the seller of record, banks want to retain visibility and control over funding instruments, and platforms want secure delegated payment handoffs that are verifiable end-to-end. Standards like UCP, AP2 and MCP are intended to satisfy those constraints while enabling agentic experiences to scale.
However, protocol endorsement is only the first step. Widespread adoption will require merchant integrations, platform incentives, and clear governance around liability, privacy and data minimization, areas where industry coordination and regulator engagement will matter greatly.
Security, liability and merchant implications
Mastercard emphasizes that agentic commerce must be secure and auditable by design: agent registration, verification, tokenized credentials and transaction trace logs form part of the acceptance frameworks that Mastercard proposes to reduce fraud and clarify accountability. Those mechanisms are intended to let issuers and merchants detect anomalous agent behavior and to support dispute resolution.
For merchants, agentic commerce presents opportunities and risks. On the upside, AI agents can increase conversion by reducing friction and personalizing offers; on the downside, merchants must integrate with new message formats, honor delegated discounts or loyalty credentials, and ensure fulfillment and returns align with agent-driven orders.
Banks and regulators will watch closely: delegated payments change the attribution of consent and may require clearer rules about consumer authorization, revocation, and the responsibilities of agents, platforms and networks when errors or abuse occur. Industry standards aim to make these responsibilities technically enforceable.
What comes next and how companies should prepare
The next 12, 24 months are likely to bring a mix of accelerated integrations and continued experimentation. Mastercard’s Agent Pay, platform toolkits and acceptance framework, combined with Google’s UCP and endorsements from wallets and retailers, create a scaffolding that could move agentic commerce from pilots into mainstream use cases.
Companies that want to prepare should inventory their checkout flows, ensure tokenization and passkey support, adopt machine-readable API documentation where available, and participate in pilot programs so they can shape consent models and liability rules. Merchants should also evaluate how they will surface agent-driven commerce offers while maintaining merchant-of-record responsibilities.
Finally, cross-industry governance, covering identity, consent, auditability and dispute handling, will be decisive. Organizations that engage with emerging standards now can influence how agentic commerce balances convenience with consumer protection.
Mastercard’s push to standardize agentic commerce is a key example of how networks, platforms and standards bodies are working in parallel to make AI-driven shopping interoperable and secure. By combining Agent Pay, developer tooling and acceptance frameworks with broader protocol work like Google’s UCP, the industry is trying to prevent fragmentation and to embed trust mechanisms in the very fabric of agentic transactions.
Whether agentic commerce reaches mass adoption will depend on merchant uptake, clear governance and user confidence that agents act transparently and reversibly. If the technical and regulatory building blocks are put in place, consumers and businesses could soon benefit from AI that not only recommends, but also reliably and safely completes purchases on their behalf.
