Agentic Payments

Protocols and standards for AI agents that execute payments on behalf of users—Mastercard Agent Pay, ACP, AP2, x402, and more.

Agentic payments are transactions completed by AI agents on behalf of customers, with security and consent built in. In early 2026, Banco Santander and Mastercard completed Europe’s first live end-to-end payment executed by an AI agent, processed through Santander’s live payments infrastructure and Mastercard Agent Pay. This page tracks the main protocols and frameworks shaping agentic payments.

Agentic payment protocols

Card network / frameworkProprietary

Mastercard Agent Pay

Mastercard

Enables secure, authenticated agentic transactions where AI agents act on behalf of consumers. Transactions flow through existing card rails; issuers, acquirers, and merchants can recognize that an agent conducted the transaction. Pilots with Santander (Europe’s first live agentic transaction) and in Australia (Commonwealth Bank, Westpac).

Focus

Card-based agentic commerce, regulated bank framework

Status: Pilot / expanding

Open standardApache 2.0

Agentic Commerce Protocol (ACP)

OpenAI + Stripe

Open-source protocol (Apache 2.0) for AI-powered commerce. Powers Instant Checkout in ChatGPT so users can buy from merchants in chat. Merchants keep customer relationships, decide whether to accept orders, and fulfill with existing systems. Uses REST APIs and Stripe’s Shared Payment Token as the initial delegated payment provider.

Focus

Consumer checkout in chat, merchant control, standardized flow

Status: Live (e.g. US ChatGPT + Etsy; Shopify coming)

Open standardApache 2.0

Agent Payments Protocol (AP2)

Google

Protocol for agent-to-agent and multi-agent commerce, with a focus on B2B and complex negotiations. Provider-agnostic and protocol-agnostic. Primary implementation via Google ADK and Gemini.

Focus

Multi-agent commerce, B2B, flexibility

Status: Emerging

Interoperability standardOpen / in development

Universal Commerce Protocol

Google

Aims to create interoperability between AI agents and merchants so agents can discover and transact with merchants in a consistent way. Part of broader agentic commerce standards collaboration.

Focus

Agent–merchant interoperability

Status: In development

Agent-to-agent standardOpen / in development

Agent2Agent Protocol

Google

Supports direct transactions and coordination between AI agents, complementing agent–merchant flows. Emphasizes clear intent, secure credentials, and verifiable agent identity.

Focus

Agent-to-agent transactions

Status: In development

Trust / identity frameworkProprietary

Agentic Token Framework

Mastercard

Framework for trusted AI transactions: clear user intent, secure credentials, and verifiable agent identity. Helps protect merchants from fraud and strengthens authentication for issuers.

Focus

Trust, identity, authentication for agentic transactions

Status: Rolling out with Agent Pay

Open payment protocolApache 2.0

x402

Coinbase (open protocol)

Uses the HTTP 402 Payment Required status code to enable instant stablecoin payments over HTTP. Lets APIs, microservices, and AI agents pay per request without accounts or sessions. Supports micropayments and machine-to-machine payments. Coinbase offers a hosted facilitator; protocol is Apache 2.0 with TypeScript and Go SDKs.

Focus

Stablecoin payments over HTTP, API monetization, AI agent payments

Status: Live (Base, Solana; Coinbase Payments MCP uses x402)

Industry standard effortOpen

Open standard for agentic commerce

Stripe (collaboration)

Stripe is collaborating on an open standard for agentic commerce alongside partners, with a focus on safe, interoperable AI-driven checkout and delegated payment flows.

Focus

Open standard, interoperability, safety

Status: In development

Protocol comparison

ProtocolStewardFocusLicense
Mastercard Agent PayMastercardCard-based agentic commerce, regulated bank frameworkProprietary
Agentic Commerce Protocol (ACP)OpenAI + StripeConsumer checkout in chat, merchant control, standardized flowApache 2.0
Agent Payments Protocol (AP2)GoogleMulti-agent commerce, B2B, flexibilityApache 2.0
Universal Commerce ProtocolGoogleAgent–merchant interoperabilityOpen / in development
Agent2Agent ProtocolGoogleAgent-to-agent transactionsOpen / in development
Agentic Token FrameworkMastercardTrust, identity, authentication for agentic transactionsProprietary
x402Coinbase (open protocol)Stablecoin payments over HTTP, API monetization, AI agent paymentsApache 2.0
Open standard for agentic commerceStripe (collaboration)Open standard, interoperability, safetyOpen

Frequently Asked Questions

Agentic payments are transactions initiated or completed by AI agents on behalf of a user. The agent acts with the user’s consent and within defined rules, using secure credentials and identity so issuers, acquirers, and merchants can recognize and trust the agent-driven flow.

Mastercard Agent Pay is a framework that lets AI agents complete card-based payments on behalf of consumers. Transactions run over existing card rails; Santander and Mastercard completed Europe’s first live agentic transaction, and Australia has seen pilots with Commonwealth Bank and Westpac. The solution is still in pilot and not yet fully commercial.

The Agentic Commerce Protocol (ACP) is an open-source standard (Apache 2.0) developed by OpenAI and Stripe for AI-powered commerce. It powers features like Instant Checkout in ChatGPT, so users can buy from merchants inside chat. Merchants keep control of orders and fulfillment; Stripe’s Shared Payment Token is the initial delegated payment provider.

x402 is an open payment protocol that uses the HTTP 402 Payment Required status code to enable instant stablecoin payments over HTTP. It allows APIs, services, and AI agents to pay per use without accounts or sessions. Coinbase supports it via a hosted facilitator and Payments MCP; the protocol is Apache 2.0 with TypeScript and Go SDKs.

Static PSD2 consent is being extended with policy-based authorisation, explicit delegation and continuous validation. See the Consent and trust section on this page and the Konsentus white paper for detail.

Consent and trust in agentic Open Banking

Agentic AI is changing how Open Banking services are executed: agents monitor, decide and act on behalf of users with limited continuous human involvement. That challenges static PSD2 consent models, which assume user-initiated activity and direct TPP execution. Emerging approaches extend consent through policy-based authorisation, explicit delegation and continuous validation so that consent chains—from consumer through TPP to agents and sub-agents—remain visible and constrained while preserving regulatory alignment.

Konsentus’ white paper on Agentic AI and the evolution of consent in Open Banking explores how agentic consent models can support secure, accountable agentic execution and why Open Banking should be treated as a living trust framework that adapts as technology and execution models evolve.

Related

Banks and platforms that let AI agents access accounts and payments (MCP, x402). More on consent and trust above.

Want to integrate Accounting & ERP data?

Connect to 30+ accounting platforms and ERPs through a single unified API with Apideck.