Live Agentic Checkout on Direct-to-Consumer Storefronts
PlatformDTC Research
Abstract
Of 58 well-known direct-to-consumer brands, 31 serve a live Universal Commerce Protocol discovery document, and every one answers a Model Context Protocol request with thirteen working checkout tools. An agent can transact with over half the sample today. Most of those brands did not build it.
Key Findings
31 of 58 brands (53%) serve a live UCP discovery document
All 31 answer a tools/list request with the same thirteen schema-d checkout tools, covering the full cart and checkout lifecycle
Identical protocol version, tool count and registered payment handlers across all 31 — one platform deployment, not 31 integrations
Two stores serving the same platform's assets return 404 on the discovery path; both run custom or headless front-ends
Calling the Endpoints
Two requests establish the result, and both can be repeated from a terminal against any host in the published dataset. The first returns a merchant profile: a declared protocol version, declared services, shopping checkout and fulfillment capabilities, and registered payment handlers including a per-merchant wallet identifier. The second returns thirteen tools with full JSON schemas covering the complete lifecycle — creating, updating, completing, cancelling and reading a checkout, and reading a cart. Discovery is open; execution is gated on agent identity.
Uniformity Is the Finding
Identical version, identical tool count, identical handler set across all 31 hosts. Thirty-one independent teams do not converge on a byte-identical surface. This is one platform's deployment, shipped to a merchant base that largely does not know it is there. That reframes the readiness question: for a large share of the category, agent commerce is not a roadmap item a merchant chooses, it is a property of the platform they already bought.
The Exceptions Teach the Most
Two stores in the sample serve their platform's assets but return 404 on the discovery path, and both look like custom or headless front-ends. The likely explanation is not that the platform withheld the capability but that a bespoke front-end at the customer domain never proxies the paths the platform serves automatically. For anyone who has gone headless, that is the operational warning: your platform may have shipped an agent commerce API on your behalf, and your own front-end may be the thing hiding it.
Scope and Honesty About Our Own Position
This study measures other organisations' infrastructure, which makes precision about our own obligatory. PlatformDTC storefronts do not currently serve these discovery or agentic-checkout endpoints. The study is how the gap was identified, and naming it is preferable to implying a parity that has not shipped.
Conclusion
The agentic commerce conversation is dominated by announcements about future support. Measured directly, a majority of a well-known DTC sample is already transactable by an agent. The gap between what the category believes about itself and what its endpoints actually answer is the most useful finding here, and it is reproducible with two curl commands.
Published In Full
Read the complete study, method and datasetReferences
- [1]PlatformDTC Research (2026). Agentic Checkout Is Already Live on Half of DTC. We Called the Endpoints. https://platformdtc.com/blog/ucp-live-on-half-of-dtc
- [2]PlatformDTC Research (2026). We Re-Ran the Study on 147 Brands. https://platformdtc.com/blog/147-brands-agent-commerce