AI agents in e-commerce: Shopify opens checkout

Shopify has opened checkout to browser-based AI agents: they can now move beyond product search and cart building to completing an order. The important limit is buyer authorization, so the agent is not being handed a blank check to purchase on its own.

For store owners, the message is practical: the buying journey is starting to change. Some customers may arrive through a personal agent that compares options, checks delivery choices, and helps them approve the final purchase.

What exactly did Shopify open to AI agents?

According to TechCrunch AI, Shopify announced that browser-based AI agents can now complete purchases on Shopify merchants’ sites. Before this update, Shopify’s WebMCP support for storefronts and carts let agents browse a merchant’s inventory, search for products, and add items to a cart.

The new support extends WebMCP to checkout, including Shop Pay. That means an agent can read the checkout screen, update it, and submit the transaction after the buyer authorizes it, without depending on screenshots or scraping web pages.

The update introduces three tools: get_checkout, update_checkout, and complete_checkout. They let an agent inspect a checkout, change details such as the customer’s address or delivery option, and then place the order once the buyer has authorized it.

Shopify said the feature is rolling out to all eligible Shopify merchants. That is a different direction from some retailers, including Amazon and, as TechCrunch notes, Adidas, which have been blocking AI agents from making purchases on users’ behalf on their platforms.

Why does WebMCP matter for agent-based shopping?

The point of WebMCP is that an agent running in the buyer’s browser does not need to behave like a person clicking around a page. It receives structured tools for specific tasks: inspect checkout, update fields, and complete the purchase.

That is different from a rougher approach where an agent tries to understand the interface from an image or extracts data from HTML designed for humans. Shopify presents WebMCP as a more efficient path for agents operating inside the buyer’s browser.

Shopify already offers a hosted Model Context Protocol, or MCP, server for server-to-server work. WebMCP addresses a different situation: the agent is on the buyer’s side, inside the browser. Both approaches use Shopify’s Universal Commerce Protocol, UCP, which provides a common way to search for and discover products, build carts, and check out.

• the agent can work with more accurate facts about products and orders;
• the merchant depends less on how an agent reads a human-facing interface;
• required disclosures and handoff requirements can be handled in the flow;
• checkout remains tied to buyer authorization.

What does this mean for a business with its own AI agent?

If a company already has a personal AI agent for customers or employees, this news points to where the market is going: agents are getting access not only to information, but also to actions. Buying, booking, submitting a request, or changing a delivery address are no longer purely manual workflows.

For a small business, the important part is not only Shopify support. The broader lesson is that your processes need to be understandable not just to people, but to agents. If an agent cannot reliably understand order status, delivery terms, price, address, or the point of consent, automation will keep falling back to manual checks.

The practical takeaway for anyone considering their own AI agent is to start with an action map, not with a pretty chat window. At NekoAgent, this is one of the first boundaries we look at: where the agent only advises, where it prepares an action, and where it can execute after explicit human confirmation.

• List low-risk actions the agent may perform: find a file, check a status, prepare a draft.
• Mark actions that require confirmation: send an order, change customer data, make a payment.
• Check whether your data is structured, not only presented as pages and spreadsheets for human eyes.
• Decide in advance how the agent will show the final summary before the last step.

Where are the risks for store owners?

The main risk is not that an agent clicks faster than a person. The risk is that a business fails to separate preparing an action from performing it, and the customer does not clearly see what they are authorizing.

Shopify’s description of the update emphasizes buyer authorization. For businesses, that should become a basic rule: the agent may help assemble the order, but final consent must be clear, checkable, and tied to a specific action.

• Show the customer the order summary before confirmation: product, address, delivery option, and payment.
• Do not blur the line between the agent’s recommendation and the buyer’s final consent.
• Make sure required terms and disclosures are not lost inside the agent conversation.
• Pay extra attention to scenarios where an agent changes the address or delivery method.

There is also a strategic risk: if buyers increasingly delegate discovery to agents, a store that describes products and terms poorly may become less visible to those assistants. A polished banner will not help an agent if it cannot reliably retrieve facts about the product.

How should businesses prepare for agentic commerce?

Not every business needs to rebuild its store around WebMCP right away. But every business should check whether its sales process is ready for a world where part of the decision is prepared by an AI agent.

• Collect the common questions buyers ask before checkout: delivery, returns, what is included, address, payment options.
• Turn the answers into clear, reusable information, not just marketing copy on a page.
• Separate product data, cart data, and order data so an agent can understand what is description, what is selected, and what is a final action.
• Set a confirmation rule: any action involving money, an order, or personal data needs explicit human approval.
• Review internal processes: if a manager has to search five places for an answer, an agent will also struggle.

Shopify’s move does not mean agents will make every purchase tomorrow. It does show that a major e-commerce platform is preparing infrastructure where an agent is a normal participant in the transaction, not an obstacle to block.

Source: TechCrunch AI: Shopify opens checkout to browser-based AI agents

Quick answers