What Is Agentic Commerce? How AI Agents Shop and Pay (2026)

Agentic commerce is when an AI agent shops and pays for you. The agent finds the product, decides what to buy, and completes the payment, with little or no human input beyond the rules you set. It is one of the fastest moving parts of the agent economy, with Visa, Mastercard, OpenAI, Google, and Stripe all building for it. McKinsey expects up to $1 trillion in US retail to run through agentic commerce by 2030.

This guide explains how agents pay, the main protocols, the companies involved, and the risks. It is part of our look at the agentic AI landscape, and it builds on recent funding stories like Pliant opening B2B payments to agents.

What is agentic commerce?

Agentic commerce is a form of online shopping where an AI agent handles the buying for you. Instead of recommending options for you to click, the agent searches, compares, decides, and pays on its own, within limits you set. A human may still set up the rules, check in now and then, or approve big purchases. The difference from older tools is simple. A recommendation engine suggests. An agent acts.

How do AI agents pay for things?

AI agents pay using special payment credentials that carry built in limits. Instead of handing over your real card, the agent gets a single use token that is capped at a set amount and expires quickly. The agent also carries proof that you gave it permission to buy. This keeps control with the user. The agent can only spend what the rules allow, and every payment can be traced back to your approval. The exact method depends on which payment protocol the agent uses.

What Is Agentic Commerce? How AI Agents Shop and Pay

What are the main AI agent payment protocols?

There are four main AI agent payment protocols in 2026. Each solves a different part of the problem.

ProtocolBackersWhat it doesBest for
ACP (Agentic Commerce Protocol)OpenAI, StripeStandardizes checkout between agents and merchantsShopping agents buying goods
AP2 (Agent Payments Protocol)Google, 60+ partnersProves an agent has permission to payEnterprise, auditable authorization
x402Coinbase, CloudflareInstant stablecoin payments over the webMachine to machine payments
MPP (Machine Payments Protocol)Stripe, TempoSession based micropayments across railsHigh frequency micropayments

ACP handles the checkout itself, using single use payment tokens. AP2 is the trust layer, using signed mandates that record what you authorized. x402 moves stablecoins between machines for things like API calls. MPP lets an agent pre authorize a budget, then stream small payments inside it. Many real systems combine them.

What are Visa Intelligent Commerce and Mastercard Agent Pay?

Visa Intelligent Commerce and Mastercard Agent Pay are the card networks’ programs for agent payments. They let an AI agent pay with a tokenized version of your card, with spending controls and the networks’ fraud systems behind it. Visa has extended its rails to agent protocols, and Mastercard is a partner in Google’s AP2. Their role matters, because most real world buying still runs on cards, and the networks bring trust, dispute handling, and scale that newer rails do not have yet.

Which companies are building agentic commerce?

The biggest names in tech and payments are all building agentic commerce. OpenAI added checkout inside ChatGPT with its ACP standard. Google launched AP2 and is bringing shopping into Gemini with Walmart. Stripe and PayPal are building the payment plumbing. Amazon added an Auto Buy feature to its Rufus assistant. In China, Alipay, Alibaba, and Tencent are building their own agent shopping and payment tools. Startups are filling the gaps too, like Pliant in B2B payments and firmly in merchant enablement.

How big is the agentic commerce market?

The agentic commerce market is forecast to be huge. McKinsey projects up to $1 trillion in US retail revenue running through agentic commerce by 2030, and $3 trillion to $5 trillion globally. Early signs back the trend. Adobe reported that AI referred traffic to US retail sites rose nearly 400 percent year over year in early 2026. Gartner expects 40 percent of enterprise apps to include task specific agents by the end of 2026, up from under 5 percent in 2025.

What are the risks of agentic commerce?

The main risks are liability, disputes, and security. If an agent buys the wrong thing or is tricked into a bad purchase, it is not yet clear who is responsible. Dispute resolution and agent authentication standards are still being built. There are privacy concerns too, since agents need access to payment and personal data, and that has slowed adoption in the US. Security is the deepest worry. An agent that can spend money is a new target, so prompt injection and fraud have higher stakes here than with a chatbot.

How can merchants prepare for agentic commerce?

Merchants can prepare by making their catalogs machine readable and picking which protocols to support. Agents need clean, structured product data, with real time pricing and stock, to buy with confidence. From there, a merchant can connect to a checkout standard like ACP, or use a platform that handles the integrations. Getting this right early matters, because AI driven traffic to retail sites is rising fast, and stores that agents cannot read will be skipped.

Frequently asked questions

What is agentic commerce?

Agentic commerce is online shopping where an AI agent searches, decides, and pays for a purchase on your behalf, within limits you set. It goes beyond recommending products to actually completing the buy.

How do AI agents pay for purchases?

AI agents pay with single use, capped payment tokens rather than your real card, and they carry proof that you authorized the purchase. The exact method depends on the payment protocol, such as ACP, AP2, or x402.


What is the Agentic Commerce Protocol (ACP)?

ACP is an open standard from OpenAI and Stripe that standardizes checkout between AI agents and merchants. It uses single use payment tokens and is supported by companies like Shopify and PayPal.

Is agentic commerce safe?

It is built with controls like spending caps and authorization proof, but risks remain around liability, disputes, fraud, and privacy. Letting software spend money raises the stakes, so security is the biggest open question.

How big will agentic commerce be?

McKinsey projects up to $1 trillion in US retail and $3 trillion to $5 trillion globally running through agentic commerce by 2030.

Scroll to Top