Shopify Just Let Browser AI Agents Complete Your Checkout. There Is No Setting to Turn On, and None Documented to Turn Off.

Shopify posted “WebMCP support for checkout” to its developer changelog on September 28. Browser-based AI agents can now read and update a Shopify checkout and submit it once the buyer confirms. Shopify says it does not require merchant configuration. Your checkout is now something an agent can operate, not just something a human clicks through.

The recommendation

Do not turn anything on, because there is nothing to turn on. Spend one hour this week on the four places where an agent-driven checkout can cost you money or data: discount codes, blocking checkout extensions, bot rules in front of your store, and order attribution. Then run one test order through an agent-capable browser so you see what your checkout looks like from the agent’s side. If your brand has channel or pricing restrictions that make agent purchases a problem, ask your Shopify rep now whether a control exists, because the documentation does not describe one.

What Shopify actually shipped

Per Shopify’s developer documentation, Checkout WebMCP registers four tools in the buyer’s browser on eligible checkouts:

• get_checkout reads the current checkout state, messages, and post-completion order details.
• update_checkout changes buyer contact details, fulfillment, discount codes, declared fields, and payment.
• complete_checkout places the order after the buyer confirms, or opens a configured review step where the buyer authorizes submission.
• navigate_to_storefront sends the buyer back to the storefront without placing an order or changing the cart.

Three details matter for operators.

First, the buyer stays in the loop. Shopify’s docs instruct agents to show the buyer the current order and total and get permission before placing it. The tools hand control back to the buyer for 3D Secure and for blocking UI extensions. Only a completed status confirms an order.

Second, payment is constrained. Shop Pay customers can use saved cards. Guest checkouts can use Shop Pay approvals or billing addresses. Shopify states that Checkout WebMCP does not accept new card details.

Third, agents are supposed to identify themselves. Requests must carry Web Bot Auth headers, and Shopify’s docs say that without them, bot detection might deprioritize or block the request. Shopify named Meta’s Muse and a partner called Instinct as agents working with the feature, per TechCrunch’s report on the launch.

What the documentation does not tell you: which checkouts are “eligible,” whether a merchant can switch this off, and how agent-driven orders show up in attribution. The exclusions list lives on a separate page. Treat those three as open questions until Shopify answers them, not as good news.

Why this is a revenue issue, not a tech curiosity

An agent that can edit your checkout is a very fast, very literal shopper. It will enter every discount code it can find. update_checkout accepts a codes array of buyer-entered codes, and automatic discounts stay applied. Agents scan the open web for coupons the same way browser extensions do.

Here is the illustrative math. These numbers are not from Shopify, they are a stress test you can rerun with your own data:

• 20,000 orders a month, $70 average order value.
• If agent-assisted checkouts reach 3% of orders, that is 600 orders.
• If a forgotten 20% code (an old influencer code, a leaked email-capture code, a win-back code that never expired) lands on half of those, that is 300 orders at $14 off, or $4,200 a month in margin you did not plan to give away.
• At 10% agent share, the same leak is $14,000 a month.

Humans stack and hunt codes too. The difference is speed and consistency. An agent tries the code every time, on every order, with no hesitation.

Your audit, in order of dollar risk

1. Discount codes. Export every active code in Shopify admin. Kill anything expired in spirit: old creator codes, one-time campaign codes, codes attached to a partner that no longer exists. Confirm usage limits and one-per-customer rules on the rest. Confirm which codes combine with automatic discounts, because those stay applied when an agent edits the code list.

2. Blocking checkout extensions. Shopify says the tools hand control back to the buyer when a blocking UI extension needs input. Any blocking extension in your checkout (age verification, a mandatory upsell step, a required consent field, a custom form) now interrupts the agent flow. That can be exactly what you want for compliance. It can also kill conversions from agent-assisted buyers for no reason. List every extension in your checkout and mark each one as required or nice to have. Remove the second kind from the blocking path.

3. Bot rules in front of your store. If you run a WAF, a bot-management app, or aggressive rate limits on checkout paths, test them. Shopify identifies verified agents through Web Bot Auth and says unverified requests may be deprioritized or blocked. What your own third-party layer does with a verified agent is your problem to test, not Shopify’s. Ask your security vendor or app provider directly. Do not assume a rule written for scrapers and card testers treats a legitimate agent correctly.

4. Attribution. The documentation does not address how agent-assisted orders are labeled in Shopify Analytics, UTMs, or your pixels. That is a gap. Place a test order through an agent-capable browser, then look at what landed in the order source, the customer journey, and your ad platform conversion events. If agent orders show up as direct traffic, your channel mix reads wrong and any “direct is growing” story you tell in Q4 needs a footnote. Decide now how your team will tag or annotate these orders, so you are not reconstructing it in January.

5. Restricted products and MAP. The docs do not describe a per-product or per-store opt-out. If you sell items with age, region, or channel restrictions, or you sit at a negotiated MAP price, confirm your checkout rules enforce those at checkout, not only in storefront presentation. Enforcement at checkout works for humans and agents alike. Enforcement that lives only in page copy does not.

The strategic read

Shopify and Amazon are moving in opposite directions. Amazon and Adidas are reported to be blocking third-party AI agents from purchasing, per coverage of the Shopify launch. Shopify is opening checkout to them. If agent-assisted buying grows, DTC stores on Shopify get a distribution surface Amazon-only sellers do not have. That is a real argument for making your own checkout clean and your product data complete, which we covered in the Shopify Meta AI channel piece. It is not a reason to assume volume shows up. Agent-assisted checkout is early, and no volume figures were published with the release.

Your checklist this week

1. Export and prune active discount codes. Set usage limits on anything you keep.
2. List every blocking checkout extension. Mark each required or optional.
3. Ask your bot-protection vendor how they treat Web Bot Auth verified agents.
4. Place one test order through an agent-capable browser. Check order source, customer journey, and pixel events.
5. Write down how your team will label agent-assisted orders in reporting.
6. Email your Shopify rep: what defines an eligible checkout, and is there a merchant control? Save the answer.

If you want a second set of eyes on your checkout, discount hygiene, and agentic setup across Shopify and your marketplace channels, book a call.