Developer Changelog / Checkout × AI agents

WebMCP now supports checkout
Browser agents can help right up to payment

AI agents running in the buyer's browser can now use WebMCP tools to read and update an in-progress checkout. When the buyer's own input is required, such as 3D Secure authentication, control is handed back to the buyer.

On this page
  1. What's changing (30-second overview)
  2. How it works: how agents and buyers split the work
  3. 4 tools available in checkout
  4. The entire shopping journey is now covered
  5. Requirements and prerequisites
  6. Typical call flow (4 steps)
  7. 5 key points for developers
  8. 3 business use cases
  9. One-line summary for proposals

1What's changing

WebMCP tools for the storefront and cart were already available. Now tools for checkout have been added, so
browser agents can now help with product discovery → cart → checkout → order review from start to finish.

Before: up to the cart

Storefront and cart tools were already available, but actions after entering checkout weren't covered by the tools.

Now: up to order review

Agents can read, update, and (after the buyer confirms) submit the active checkout in the buyer's browser session.

2How it works: how agents and buyers split the work

Browser agent Runs in the buyer's browser WebMCP checkout-web WebMCP tools ⇅ Same state as the checkout UI No new APIs are exposed Buyer input required 3D Secure authentication Blocking UI extensions Hand back Buyer Operate and verify it yourself
The tools run inside checkout-webandshare the same state as the checkout UI. Instead of going through a back door separate from the screen, the agent operates the very checkout the buyer is looking at.
When the buyer's own input is required (3D Secure authentication, blocking UI extensions, etc.), the toolhands control back to the buyer.

34 tools available in checkout

Navigate to storefront
navigate_to_storefront
Return from checkout to the storefront
Read state
get_checkout
Checkout state, messages, and order details after completion
Update fields
update_checkout
Update supported checkout fields
Submit
complete_checkout
After the buyer confirms,submits the checkout
update_checkout — the specific list of "supported fields" it can update isn't given in this article. Argument and return-value specs for each tool aren't given either (see the "WebMCP tools for checkout" docs for details).

4The entire shopping journey is now covered

Find products Storefront tools Manage cart Cart tools Checkout New Order confirmation New (get_checkout) Available NEW
StageToolsStatus
Product discoveryStorefront toolsAvailable
Cart managementCart toolsAvailable
Checkoutget_checkout / update_checkout / complete_checkout / navigate_to_storefrontNew
Order confirmationget_checkout(order details after completion)New in this update

5Requirements & prerequisites

No merchant setup required

The article explicitly states that no merchant setup is required.

No new APIs are added

The tools run inside checkout-web and don't expose any new APIs.

Requires the buyer's browser

The target is an active checkout within the buyer's browser session. Intended for agents that shop through the browser.

This article doesn't cover supported plans, regions, or browsers, or how to disable (opt out of) the feature.

6Typical call flow (4 steps)

1

Read state

get_checkout to get the current state and messages.

2

Update fields

update_checkout to fill in supported fields.

3

Buyer reviews

If 3D Secure or similar is required, control returns to the buyer.

4

Submit & confirm order

After confirmation, call complete_checkout; once complete, use get_checkout for order details.

Final submission requires the buyer's confirmation.The agent isn't designed to finalize payment on its own. Whenever the buyer's own input is needed, control returns to the buyer.

75 key points for developers

STATE

1. Shares the same state as the UI

The tools run inside checkout-web and use the same state as the checkout UI. Agent updates and what's shown on screen don't maintain separate states.

2. Not a new API

No new server-side endpoints or APIs are added. Think of it as a set of tools that runs entirely within the buyer's browser.

UI ext

3. Blocking UI extensions hand control back to the buyer

If a blocking UI extension requires buyer input, control passes from the agent back to the buyer. Keep in mind that your own Checkout UI extensions become "stopping points" in agent-driven purchase flows.

3DS

4. Humans handle authentication

Identity verification such as 3D Secure authentication is designed to return to the buyer. Submission (complete_checkout) likewise happens only after the buyer confirms.

5. End-to-end together with storefront and cart

Combined with the existing storefront and cart tools, agents can assist all the way from product discovery to order confirmation. Teams building commerce agents can design experiences around the full tool set (see the "building commerce agents" docs for details).

83 practical use cases

USE CASE 1

Check whether your checkout customizations still work smoothly through agents

Challenge
Stores with Checkout UI extensions don't know at which points a purchase made through a browser agent hands control back to the buyer.
Approach
Inventory the UI extensions in your checkout and identify blocking ones that require buyer input.
Impact
Clarifies where control returns to the buyer in agent-driven purchases, so you can sort necessary extensions from unnecessary ones.
Tech notes
Blocking UI extensions return control to the buyer (per the post). The post doesn't say how an extension is classified as blocking.
WebMCP
USE CASE 2

Design commerce agents end-to-end, through order confirmation

Challenge
You want to build purchase assistance for Shopify stores with a browser agent, but haven't decided how to handle everything after the cart.
Approach
In addition to the storefront and cart tools, design the flow asget_checkout → update_checkout → buyer confirmation → complete_checkout → get_checkout(order details).
Impact
Build everything from product discovery to order confirmation with the same toolset. No merchant setup required.
Tech notes
Assume control returns to the buyer for steps like 3D Secure, and give the agent explicit "waiting" and "resuming" states.
?
USE CASE 3

Explain "purchases via AI agents" to merchants

Challenge
You can't answer questions like "Will AI make payments without my approval?" or "Do I need to configure anything?"
Approach
Summarize and share the key points of this post: no merchant setup, no new APIs, submission only after buyer confirmation, and 3D Secure and similar steps return to the buyer.
Impact
Addresses merchant concerns and lets you plan initiatives assuming agent-driven purchases can already happen.
Tech notes
The post doesn't cover how to opt out or what's in scope. Check the official documentation for the latest details before advising merchants.

9One-line summary for proposals

"Browser agents can now read and update Shopify checkout via WebMCP,
so AI can help with everything from finding products to confirming the order. No merchant setup and no new APIs,
and it's designed for safety, with authentication and final confirmation handed back to the buyer."