We checked 141 well-known storefronts for UCP. Almost every ready one was ready by accident.

Yesterday I pointed our checker at 141 storefronts. Household DTC brands, the UCP launch partners, the big-box retailers, a spread of well-known shops. One pass each, the same engine that runs our public check, arriving the way any agent arrives: anonymously, over HTTPS, asking for /.well-known/ucp.

Here is what came back.

The numbers

48 of 141 stores, just over a third, served a UCP manifest. Every one of the 141 responded to the request itself, so this is not a story about dead sites.

Now the part that made me re-run the crawl because I assumed I had a bug: 46 of those 48 manifests score exactly the same. Grade B, 92 out of 100. Same shape, same capabilities, same gaps.

That is not 46 teams independently converging on the same implementation. That is a platform. Allbirds, Gymshark, Glossier, Skims, Liquid Death, Fenty Beauty, all of them Shopify storefronts, all of them serving what is recognisably the same default manifest. Shopify co-developed UCP, and it has evidently switched it on for its merchants. Most of the agent-ready stores on the modern web are ready without anyone at the company having decided anything.

The two exceptions were Made In Cookware at a D and Fishwife at an F, which look like partial or modified setups rather than defaults. I mention them not to dunk on anyone but because they illustrate the other side of platform defaults: touch the machinery and the grade moves, in either direction.

The absentees are the interesting part

Target. Walmart. Etsy. Wayfair. The companies that stood on stage when UCP launched in January. Our anonymous fetch got no manifest from any of them.

I want to be careful here, because absence in our check does not prove absence. A manifest can exist and be served selectively. A bot rule can turn away an unfamiliar requester before the file is ever reachable. What I can say precisely is this: an agent that shows up the way our checker shows up, which is the way most agents show up, does not get a manifest from the biggest names on the launch roster.

Either it is not published at the standard path yet, or the front door is closed to strangers. Both are worth knowing if you are betting on agentic commerce, and the second one is the failure mode we bang on about most: everything configured correctly behind a door nobody can open.

WebMCP came along for the ride

46 stores also exposed WebMCP tools, and the overlap with the Shopify set is nearly total, which says the same thing: the platform shipped it. More interesting were the five stores with WebMCP tools and no UCP manifest: Zalando, Warby Parker, On Running, Chewy and iRobot. Someone at those companies made a choice, because no platform default explains them.

What I take from one afternoon of data

Three things, held loosely, because 141 hand-picked famous domains is not a random sample and a static check from one vantage point is not the whole truth.

First, "a third of stores are agent-ready" is true and misleading in the same breath. Readiness right now is overwhelmingly something that happened to merchants, not something they did. Nobody has customised anything: 46 identical grades means 46 untouched defaults.

Second, the default is a B, not an A. The platform gets you most of the way and stops. Whatever separates a B from an A on your storefront is currently yours to close, and almost nobody has started, which means the differentiated tier of agentic commerce is sitting there unclaimed.

Third, the gap between the launch announcement and the launch partners' own front doors is real, and it is exactly the kind of gap that never shows up in anyone's dashboard. Which is, not coincidentally, the reason this product exists.

We will run this again as the index grows and publish what changes. If you want to know which bucket your own store is in, the check takes a minute.