As AI agents reshape online shopping, humans are no longer the only ones making purchases. This guide explains agentic commerce and its implications for both businesses and buyers. We’ll cover key concepts you should understand, practical use cases, and how agentic commerce will evolve. You’ll also learn actionable strategies so that your business can adapt in this new and quickly changing environment.

What is agentic commerce?

Agentic commerce is a form of online shopping that involves AI agents finding, comparing, and potentially making purchases for customers. The customer tells an agent what they want (e.g., “book me the cheapest nonstop flight that lands before noon”), and the agent goes out, evaluates the options, and presents them to the customer. Once the customer approves, the agent completes the purchase on their behalf. Customers don’t click through a hundred tabs. They don’t even need to see the checkout page.

This kind of transaction flips the usual script. Instead of businesses designing every detail around human shoppers navigating their sites, they’re now being approached by AI agents acting as their customers’ proxies. And those agents want clean and structured data, fast responses, and zero guesswork.

You might see this process being called “a-commerce”—it’s referring to the same thing: a world where digital agents are no longer just helpers, but rather act as decision-makers, transaction initiators, and, increasingly, the ones making purchases.

Key concepts to know

If you’re new to agentic commerce, here are the ideas worth getting familiar with. This section won’t cover everything, but it’ll give you the mental framework to understand how this new mode of commerce actually works.

Agents

AI agents are autonomous or semiautonomous software programs that use artificial intelligence to act on a user’s behalf. What they do goes far beyond surfacing search results—they decide what to buy, when to buy it, and where. They can follow instructions, weigh trade-offs, and transact. In other words, they don’t assist with the purchase—they actually are the purchaser.

Agentic economy

As agents become capable of transacting, they stop being behind-the-scenes helpers and start becoming active participants in the economy. That changes who businesses are talking to (sometimes it’s large language models (LLMs), not a human) and what systems need to be in place. Discovery, conversion, attribution, fraud prevention—it all gets reshaped.

For AI at Stripe, we’re thinking through every contour of how these systems need to work right now and in the future.

Agent-initiated payments

Agent-initiated payments are transactions made by an AI agent for a user. When an agent makes a purchase, it needs a way to send money that’s secure, auditable, and authorized. Stripe’s payments infrastructure is designed to support these transactions. It handles authorization, confirmation, and security without breaking the agent’s flow or requiring a human to jump in mid-checkout.

Model Context Protocol (MCP)

Agents need context to act effectively. MCP, a framework that enables LLMs and AI agents to understand contextual information across systems, gives them just that—structured, machine-readable access to information like inventory, prices, and checkout logic. Instead of scraping messy web pages or simulating clicks, agents use MCP to interact directly with a business’s systems. Developers can also expose their services as monetizable MCP servers that AI agents can understand and use.

Stripe’s Model Context Protocol (MCP) allows AI applications and popular AI-powered code editors to interact with Stripe’s services via natural language.

How does agentic commerce work?

Here’s the basic idea: instead of users navigating your site, their agent does. The agent knows what the user wants, it knows what you offer, and it’s equipped to handle the entire transaction—from discovery to decision-making to checkout, only looping a human back in for approval before purchasing.

Let’s say a customer asks the following: “Find me a pair of waterproof hiking boots in a size 8 under $150 that can be delivered by Friday.” The agent scans available options within the parameters of the customer’s question, using protocols like MCP to query product data in real time. It checks return policies, reads shipping estimates, verifies availability, and places the order.

If the business in this example uses Stripe Issuing, a temporary virtual credit card number is created for the agent so it doesn’t access real card details. The payment then goes through a Stripe-verified flow designed to authorize on behalf of the user, with all the usual protections in place: fraud detection, receipts, and reconciliation-ready metadata.

While agents can navigate current sites, companies are starting to build for them—providing structured product catalogs, designing checkout flows to run without a human in the loop, and publishing agent-readable trust signal information like return policies, fulfillment guarantees, and contact information. Just as search engine optimization made content legible to search engines, agentic commerce will require making your business legible to agents.

If your business isn’t ready to accept purchases from agents, you’re going to miss out on those transactions entirely. That means building flows that can handle agent-initiated payments, respond with structured metadata, and pass whatever trust checks an agent is running before it “buys.”

Top use cases for agentic commerce

Now that we’ve covered the mechanics, let’s talk about how businesses are actually using this and where it’s going next. Maybe you’re building your own agent or adapting your checkout to accept agentic commerce and stay discoverable in a post-UI commerce environment—in any case, here are two major ways companies are getting in early.

AI buying agent: How to enable agents to send and spend money

The most obvious use case—and the one getting the most traction right now—is the AI shopping agent. Again, until now, we’ve largely been dealing with agents that recommend products, but now they can buy them too.

Stripe’s agentic payment flows are built for exactly this: agents that are allowed to spend but still need to follow secure, auditable rails to do it.

Perplexity’s AI shopping assistant

One of the most visible examples right now is Perplexity’s Stripe-powered shopping assistant. It lets users describe what they’re looking for in natural language—something like “size L dress for a summer wedding guest under $120”—and the assistant handles the search and checkout.

Monetize agent usage

Agents aren’t really endpoints for checkout as much as they’re intermediaries in a growing stack of AI decision-making. That opens up new ways to monetize beyond the transaction itself.

For example, you can build APIs for your business that others’ agents pay to access. You can prioritize certain listings for agents looking to narrow options quickly. You can even build fulfillment or support tiers designed specifically for agent expectations: same-day response, verified inventory, rich metadata, etc.

Stripe supports monetizing agent interactions and outcomes through various pricing strategies, whether you’re charging per query, by subscription, or based on successful actions. For example, a customer support application could implement a per-query model to use its agent, allowing businesses to pay only for the support interactions they use. With Stripe’s usage-based billing (UBB), companies can establish flexible pricing models that align costs with actual usage, ensuring customers only pay for what they use. Businesses can easily integrate Stripe services into their agentic workflows with Stripe’s agent toolkit and also monetize usage of their MCPs with Stripe’s MCP payments, providing real-time insights into payment lifecycles that enable agents to deliver informed, responsive interactions.

Future of agentic commerce

Stripe powers 78% of the Forbes AI 50, and more than 700 AI agent startups launched on Stripe in 2024—a number we expect to be surpassed in 2025.

Agentic commerce has the potential to rewrite the foundations of how commerce works online. As Stripe’s Head of Information and Data Science Emily Glassberg Sands mentioned in the Complex Systems podcast, “Agents don’t just change who’s at the checkout. They change who’s doing the searching, the deciding, the trusting—all of it.”

What that means in practice: the parts of commerce that used to be UX problems are becoming protocol problems. It means ensuring your product catalog is structured in a way where it’s easy for agents to discover and surface your products; optimizing checkout flows for agents, not just humans; creating risk layers with scoring; and implementing customer modeling purpose-built for agent-initiated transactions, where traditional signals no longer apply.

How to get started with agentic commerce

If you’re building for agents—or building systems that agents will interact with—start with infrastructure that treats them as first-class actors, not edge cases.

At a high level, that means:

• Make your data machine-readable. Expose product, pricing, availability, and policy data in formats agents can query directly. For SaaS businesses, building your own MCP ensures your software can be discoverable and interoperable with the new generation of AI tools.
  
• Support agent-initiated transactions. Agents need to be able to send and spend money—safely, on behalf of users. That includes programmatic authorization, audit trails, and fraud tooling that understands agent behavior.
  
• Let the right agents through. We plan to update Stripe Radar to help distinguish trusted agents from bots you don’t want anywhere near your checkout.
  

Using Stripe Payments or Stripe Billing, retailers can structure their product catalog in a way that’s accessible to agents in the future. Stripe is also building toward more refined access controls, such as giving agents permissioned access to specific products or broadcasting restocks to agents that previously visited your site. Stripe’s Optimized Checkout Suite and Link—Stripe’s accelerated checkout—also provide structured, agent-compatible flows. With Link enabled, retailers will be able to recognize the same customer across different AI channels, thereby maintaining the customer relationship and personalized experience, regardless of how they choose to shop.

Build for the customer behind the agent

It’s helpful to remember that agents won’t be replacing humans—they’ll just be acting on their behalf. This means that everything you optimize—your pricing, your checkout flow, your trust signals—is still for the end customer, even if you’re no longer talking to them directly.

In this environment, compatibility is table stakes. Agents will gravitate toward businesses that make it easy: structured data, clean APIs, transparent policies, real-time availability, and a frictionless experience. Whether you’re building an agent or adapting your checkout, Stripe can enhance your agentic workflows and ensure every interaction meets customer expectations.

For a deeper dive and to explore additional resources and use cases tailored for retail, check out this comprehensive guide on unlocking growth through agentic commerce.