Shopify’s browser-based shopping tools can now follow a buyer into checkout. In a September 28 developer changelog, the company said compatible AI agents can read and update an active checkout, then submit the order after the buyer confirms it.
That closes a gap in Shopify’s existing WebMCP tools. An agent could already search a merchant’s catalog, manage a cart and take the shopper to checkout. It now has structured tools for the transaction itself, without having to interpret the checkout page and imitate clicks.
Shopify is describing a buyer-confirmed flow in the buyer’s browser. The agent does not have standing authorization to purchase whatever it chooses, and support depends on the agent, browser and checkout being eligible.
The New Tools Act on the Checkout the Buyer Sees
Shopify’s Checkout WebMCP documentation describes three principal checkout tools:
get_checkoutreads the current checkout, including its status, totals and messages. After an order is placed, it can read the receipt on the Thank you page.update_checkoutchanges supported fields such as contact information, fulfillment choices, discount codes and payment selections. It does not place an order.complete_checkoutsubmits the order after the buyer confirms it.
A fourth tool, navigate_to_storefront, can return the tab to the shop when the merchant has an online storefront. The tools operate on the checkout open in that browser tab, using the same state as the interface the buyer sees. They run inside Shopify’s checkout system; they are not a new merchant-facing checkout API.
Suppose a shopper asks an agent to apply a discount and choose delivery. The agent can read the checkout, submit the supported changes, then read it again to check the resulting state and total. The buyer must change products on the page: Shopify says update_checkout ignores line items.
One implementation detail has real consequences for agent builders. update_checkout uses PUT-style replacement semantics, so an agent should send the complete desired checkout state, including values it intends to keep. Shopify warns that most omitted values can be cleared, subject to field-specific rules. Its guidance is to start each update with a fresh get_checkout result instead of constructing a partial change from an old snapshot.
The tools also expose structured status and error information. An agent can respond to the checkout’s actual state instead of assuming a button click worked or inferring success from a screenshot. After submission, it can check the receipt; a call to complete_checkout alone is not proof that an order exists.
Confirmation Comes Before Order Submission
Shopify’s stated sequence puts the buyer before the final tool call: the buyer confirms the order, then the agent calls complete_checkout. A checkout marked ready_for_complete is ready for an attempt at completion, but Shopify’s documentation explicitly says buyer confirmation is still required.
Some interactions remain with the buyer. Shopify says the tools hand control back when buyer input is needed, including 3D Secure authentication or a blocking checkout UI extension. The buyer handles page interactions such as Shop Pay login and payment challenges. For a Shop Pay buyer, an agent may be able to work with saved payment options exposed by the checkout tools, but those handoffs still apply.
So the phrase “AI agents can complete purchases” needs care. An agent may prepare checkout fields and make the final submission call; the documented flow does not authorize an unattended purchase. A sensible implementation would present the current order details and total for confirmation immediately before submission, particularly if shipping, discounts or payment choices have changed.
Shopify also tells developers not to work around a checkout tool by operating page controls themselves. If a required interaction blocks the agent, the intended path is to let the buyer handle it.
Availability Depends on the Checkout and the Agent
Shopify says checkout WebMCP requires no merchant configuration. That does not mean every shopper can bring any AI assistant to any Shopify checkout today. The tools register on eligible checkouts, and an agent must support the browser-based tooling to discover and call them. TechCrunch’s report on the rollout describes availability for eligible Shopify merchants and notes Shop Pay support.
The earlier storefront WebMCP tools cover catalog search, product information and cart management. Shopify says support for agents using WebMCP is currently limited to Chromium-based browsers. A storefront agent can call proceed_to_checkout, but the available tools change when the shopper enters checkout. Developers must refresh the tool list and check whether checkout tools are registered.
If they are unavailable, Shopify’s guidance is to hand checkout back to the buyer on the page. Managing the cart does not guarantee an agent can finish the purchase.
For merchants, this is narrower than installing or switching on an “AI checkout” feature. Shopify supplies the tools on supported surfaces without a separate merchant configuration step. In practice, a buyer needs a compatible agent and an eligible checkout. Merchants evaluating the experience should pay attention to what information the agent can read and change, and where their checkout requires a person to act.
Browser WebMCP Is Not Shopify’s Server-Side Checkout MCP
Shopify has another route for commerce agents, called Checkout MCP. The similar names conceal a difference in where the agent operates.
Checkout WebMCP serves an agent acting in the buyer’s browser. It discovers tools registered by the checkout page, and each call applies to the checkout open in that tab. Checkout MCP is the route Shopify directs developers toward when their agent runs on a server. Shopify says the two approaches share the underlying checkout object, statuses and messages. Browser WebMCP exposes that capability through browser-registered tools instead of server-side JSON-RPC.
A remote-service agent should not treat this announcement as a general-purpose API for reaching any shopper’s open tab. A browser agent shopping alongside a buyer, meanwhile, can use structured tools to understand a supported checkout without relying on screen-reading and simulated clicks. The choice starts with where the agent runs.
Checkout WebMCP implements Shopify’s UCP checkout capability, giving agents a structured representation of the transaction. That common data model does not erase the operational difference between a live browser session and a server-to-server integration.
Frequently Asked Questions
4 questions
1Can a Shopify AI Agent Place an Order Without Buyer Confirmation?
No. Shopify’s documented Checkout WebMCP flow requires the buyer to confirm the order before the agent calls complete_checkout. The agent can read the live checkout and update supported fields, but interactions such as Shop Pay login, payment challenges and blocking checkout UI return control to the buyer.
2
Sources
- September 28 developer changelogshopify.dev
- Checkout WebMCP documentationshopify.dev
- TechCrunch’s report on the rollouttechcrunch.com
- storefront WebMCP toolsshopify.dev
