Google adds carts to UCP

Shopping agents have had a fake-easy problem. They could recommend products and even kick off a purchase, but the moment a shopper wanted a normal cart, current inventory, or loyalty pricing, the flow got brittle. That gap matters because real shopping is messy — people buy more than one thing, stock changes by the minute, and discounts depend on who you are. Google’s March 19 update to the Universal Commerce Protocol, or UCP, is basically an attempt to fix that. ### What is UCP, exactly? UCP is Google’s open standard for agentic commerce — a common way for AI surfaces, merchants, and payment systems to talk to each other during shopping and checkout. Google introduced it in January 2026 with partners including Shopify, Mastercard, and Visa, and positioned it as the plumbing for direct purchases inside places like Google Search’s AI Mode and Gemini. ### What was missing before? The first version was useful, but narrow. It focused on direct buying and single-item checkout flows, which is fine for “buy this now” but not for how most people actually shop. If an agent could only pass one item through a checkout session, it still wasn’t handling the core retail behavior of building a basket, checking options, and then deciding. (developers.googleblog.com) ### So what changed on March 19? Google added three big capabilities. First, UCP can now save multiple items to a cart at once. Second, agents can query retailer catalogs for real-time product details like pricing and inventory. Third, UCP now supports identity linking, so a shopper’s loyalty relationship can carry into the transaction instead of getting lost between surfaces. Google also said it is simplifying onboarding through Merchant Center. (blog.google) ### Why do carts matter so much? Because a cart is where shopping stops being a demo and starts being commerce. A single-item purchase is the easy version of the trick. A real cart has quantities, substitutions, availability changes, shipping thresholds, promo logic, and edge cases everywhere. Adding cart support means UCP is moving closer to the shape of an actual retailer checkout, not just a one-click handoff. (blog.google) ### Why are live catalogs a bigger deal than they sound? Because stale product data breaks trust fast. If an agent recommends a size that just sold out, or quotes a price that changed ten minutes ago, the whole experience feels unreliable. Live catalog queries let an AI system ask the merchant for the current answer instead of relying on an old feed. That turns catalogs from background metadata into live transaction infrastructure. (blog.google) ### And what does loyalty linking unlock? It fixes one of the most annoying gaps in off-site shopping — losing member pricing, points, or perks when you leave the merchant’s own app or site. With identity linking, UCP can recognize that the shopper is also a loyalty member and keep those benefits in play. That matters for conversion, but also for merchant buy-in, because retailers do not want AI intermediaries stripping out the economics of their loyalty programs. (blog.google) ### What’s the catch for merchants? The hard part is no longer just exposing a checkout endpoint. Merchants now need cleaner coordination across catalog, pricing, inventory, identity, and checkout systems. In other words, the AI can only be as good as the underlying commerce stack. If those systems disagree, the agent will surface the mess instantly. That is my inference from the new feature set and Google’s implementation guidance — the protocol is getting more capable, but it also demands more operational discipline. (blog.google) ### Bottom line This update does not mean AI shopping is solved. But it does mean Google is pushing UCP past the toy stage. The bottleneck is shifting from “can an agent talk to checkout?” to “can a merchant keep pricing, stock, carts, and loyalty coherent enough for an agent to act on them?” (blog.google) (developers.google.com)