Shopify turned on WebMCP for every store
it's live for every store
About two weeks back, I covered Chrome's new Agentic Browsing audit, the panel that scores how well an AI agent can actually use a page. WebMCP was one of four checks, the newest and least settled of the bunch. Read it here →
Shopify turned WebMCP on for every store. Ilya Grigorik (Shopify's engineer working on this standard) shared that ChatGPT's built-in browser now speaks WebMCP too.
What WebMCP is
WebMCP stands for Web Model Context Protocol. MCP itself is the naming convention AI tools already use for calling functions and WebMCP borrows that idea for the browser.
It's a browser standard - your page hands an agent a menu of exact actions it can take, instead of making it read the page and click around as a person would.
On a Shopify store, that's search_catalog, get_product, update_cart, proceed_to_checkout, and six more. Shopify pre-registered all ten on your storefront.
One limit is that only Chromium-based browsers can call these tools right now. That's a question about which agents support it, not something we need to fix.
Not same as an MCP server or the Catalog API
An MCP server is infrastructure. You can build it, host it and hand it credentials.
WebMCP tools live inside the page itself and run in the shopper's own browser session. The cart it touches is their cart. The checkout it starts uses their login.
Shopify's Catalog API is different: cross-merchant discovery, an agent searching across every store on Shopify at once. WebMCP stays inside one store, on one live session.
What it changes for you
Nothing, on the setup side. Your theme is out of the loop entirely.
What an agent actually receives is the payload, the structured data behind your page, not the page itself. That's what the Everlane test below shows.
Ran it on Everlane's product page
I called the WebMCP tools on one of their products to see what came through.
Colour isn't a product option there, the kind you'd set up as a selectable option in Shopify admin. It's a word sitting in the title, and a tag. Fabric is a tag too. Alongside them sat a handful of internal-sounding tags: Selling fast, Healthy Stock, Discounted.
None of that is wrong. It's readable, though. An AI agent can find these words, but it can't filter by them the way it filters a structured option.
What to fix on your own store
Fix these four:
Colour and size need to be real options, not words in a title or a tag
Sort your tags, and drop the internal ones
Check your declared options against what's buyable
Put fit, occasion, and care into the description as plain facts
Turning colour and size into real options carries the most weight of the four. Across a sample of 400,000+ live Shopify products, we found 51 percent of what a shopper asks about sits somewhere in readable text. Only 6 percent of that is filterable.
Colour and size sitting in a tag or a title instead of a selectable one follows that same pattern - readable in text, not filterable as a field.
Pull the payload for your best seller and read it next to the product page. The audit checks the same gap - filterable vs. readable, across your whole catalog.
- Ankit
If this was useful, the next one will be too.
Weekly. Free.
Unsubscribe with one click.
