Three Moves That Changed the Settlement Landscape
The infrastructure for autonomous commercial settlement crossed a critical threshold in the final quarter of 2025 and the opening months of 2026. Three of the world's largest payment networks, each approaching from a different angle, converged on the same conclusion: AI agents need their own payment rails.
Stripe's Agentic Commerce Suite, launched in December 2025, provides a single integration point for businesses to sell through multiple AI agents. Visa's Intelligent Commerce protocol, also known as the Trusted Agent Protocol, establishes cryptographic verification for AI agents operating across merchant networks. Mastercard's Agent Pay leverages network tokenization to enable secure, AI-initiated transactions at scale.
These are not pilot programmes or research papers. They are live production systems processing real transactions. The settlement layer for agentic commerce is no longer theoretical; it is operational.
Stripe's Agentic Commerce Suite: One Integration, Every Agent
Stripe's Agentic Commerce Suite solves a practical problem that has constrained agentic commerce adoption: fragmentation. Every AI agent platform, whether ChatGPT, Microsoft Copilot, or Perplexity, has its own integration requirements, onboarding flows, and API specifications. A brand that wants to sell through all of them faces the same multi-platform integration burden that plagued the early days of mobile payment acceptance.
The Agentic Commerce Suite collapses this complexity into a single low-code integration. Businesses connect once to Stripe and become discoverable across every AI agent that participates in the Agentic Commerce Protocol (ACP), an open standard Stripe co-developed with OpenAI. ACP provides the shared technical language between agents and merchants, handling product discovery, transaction construction, and checkout completion within conversational interfaces.
Major brands have already adopted the suite. URBN (including Anthropologie, Free People, and Urban Outfitters), Etsy, Coach, Kate Spade, Ashley Furniture, and Halara are live. E-commerce platforms including Squarespace, Wix, WooCommerce, commercetools, and BigCommerce have integrated at the platform level, meaning their entire merchant bases gain agent visibility through a single upgrade.
Shared Payment Tokens
The critical innovation within the suite is Shared Payment Tokens (SPTs). These are single-use, time-limited, revocable credentials that allow an AI agent to process a payment using a customer's authorised payment method without ever seeing the underlying card number or bank details.
SPTs solve the trust equation that has historically blocked autonomous transactions. A human authorises their agent to make purchases up to a specified limit. The agent receives a scoped token. The token works across any merchant that accepts Stripe. When the transaction completes, the token self-destructs.
This architecture eliminates the need for the customer to be present at the point of sale. The agent acts as a fully credentialed proxy, but with cryptographic guardrails that prevent scope creep or credential reuse.
Visa Intelligent Commerce: Verifying the Agent
Visa's approach addresses a different layer of the trust stack. While Stripe focuses on enabling the transaction, Visa focuses on verifying the agent itself.
Visa Intelligent Commerce, formerly referred to as the Trusted Agent Protocol, establishes a cryptographic framework for distinguishing legitimate AI agents from malicious bots. In traditional e-commerce, fraud detection relies on behavioural signals: IP addresses, device fingerprints, mouse movement patterns. Autonomous agents exhibit none of these signals. They operate headlessly, programmatically, and at speeds that resemble automated attacks more than genuine purchases.
Visa's protocol solves this by issuing verifiable credentials to authenticated AI agents. When an agent initiates a transaction, the merchant's payment processor can validate the agent's identity and authorisation scope before the transaction is approved. This is structurally similar to how SSL certificates authenticate websites, but applied to autonomous commercial actors.
Stripe has confirmed integration between its SPTs and Visa Intelligent Commerce. When a customer authorises an agent through Stripe, the system can provision a Visa-backed agentic network token that carries both the payment credential and the agent's verified identity in a single cryptographic payload.
Mastercard Agent Pay: Tokenisation at Network Scale
Mastercard's contribution operates at the network tokenisation layer. Mastercard Agent Pay, also referred to as Agentic Tokens, extends Mastercard's existing token infrastructure to support AI-initiated transactions directly.
Network tokenisation replaces sensitive card data with unique digital tokens at the network level. This technology already processes billions of traditional transactions annually. Mastercard Agent Pay extends this capability to autonomous agents, creating a new token type specifically scoped for machine-to-machine commerce.
The distinction from Stripe's SPTs is architectural. SPTs are provisioned by the payment processor (Stripe) and work across the Stripe network. Mastercard's Agentic Tokens are provisioned at the network level and work across any processor that supports Mastercard tokenisation. Stripe has confirmed it will support both, meaning a single agentic transaction can carry an SPT wrapped in a network-level Mastercard or Visa token, creating multiple layers of cryptographic verification.
The Convergence Pattern
What makes this moment significant is not any single announcement in isolation. It is the convergence pattern. Three independent organisations, operating at different layers of the payments stack, arrived at structurally compatible solutions within months of each other.
This convergence follows a pattern observed in previous infrastructure shifts. When HTTP, HTML, and TCP/IP converged in the early 1990s, the result was the commercial web. When iOS, Android, and 4G converged in the early 2010s, the result was the mobile commerce explosion. The current convergence of protocol-level agent standards (Stripe ACP), identity verification layers (Visa Intelligent Commerce), and network-level tokenisation (Mastercard Agent Pay) is structurally identical.
The market data supports this trajectory. McKinsey's October 2025 analysis projected the retail sales opportunity from agentic commerce at $1 trillion in the United States by 2030, and up to $5 trillion globally. Industry research indicates that 81% of consumers are open to using agentic commerce tools to make purchases on their behalf. Stripe processed $1.9 trillion in total payment volume across its network in 2025, providing a concrete baseline for the transaction infrastructure that will underpin this growth.
What This Means for Brands
The practical implications for brands are immediate and structural.
Agent Readiness Is Now a Revenue Requirement
Brands that are not discoverable by AI agents are invisible to a growing segment of commercial activity. The Stripe Agentic Commerce Suite means that agent readiness is no longer an R&D project; it is a configuration step. Businesses using Squarespace, Wix, WooCommerce, or BigCommerce can activate agent visibility through their existing platform integration.
Payment Friction Has Been Eliminated
The combination of SPTs, Visa credentials, and Mastercard network tokens means that the payment layer is no longer a bottleneck for autonomous transactions. An AI agent can discover a product, verify the merchant, construct a transaction, authenticate the payment, and complete settlement without any human interaction or page load.
Fraud Detection Must Evolve
Traditional fraud detection systems will need to adapt to agentic transaction patterns. Visa's Intelligent Commerce protocol provides the verification layer, but individual merchants must update their fraud rules to recognise and accept authenticated agent transactions. Blocking all headless, high-speed transactions will increasingly mean blocking legitimate revenue.
Product Data Must Be Machine-Readable
Agent discoverability depends entirely on structured product data. Agents cannot browse visual storefronts or interpret marketing photography. They require JSON-LD schemas, structured catalogue data, and clean API endpoints. Brands with rich, accurate product data will be prioritised by agents; those without it will be systematically excluded.
The Five Levels of Agentic Commerce
Stripe's co-founders, Patrick and John Collison, have outlined a five-level maturity model for agentic commerce, analogous to the levels of autonomous driving. The industry is currently operating between levels one and two.
Level 1 involves AI agents that assist human shoppers with recommendations and comparison. Level 2 adds structured checkout capabilities, where the agent can construct and present a transaction for human approval. Level 3 introduces delegated purchasing within defined parameters. Level 4 enables fully autonomous procurement with self-adjusting budgets. Level 5 represents complete agent-to-agent commerce, where both buyer and seller sides are operated by autonomous systems.
The infrastructure announced by Stripe, Visa, and Mastercard collectively enables everything up to Level 3 today, with clear technical pathways to Levels 4 and 5. The limiting factor is no longer technology. It is consumer trust, regulatory clarity, and organisational readiness.
The Window Is Measured in Months
The brands that moved early on mobile commerce captured disproportionate market share that persists to this day. The same dynamic is unfolding with agentic commerce. The infrastructure is live. The payment rails are operational. The major e-commerce platforms have already integrated.
The question is no longer whether agentic commerce will happen. It is whether your brand's data, payment infrastructure, and product catalogue will be ready when the agents come looking.
