What is UCP? Universal Commerce Protocol Explained
Google and Shopify announced the Universal Commerce Protocol (UCP) at the National Retail Federation’s Big Show on January 11, 2026, with Sundar Pichai using the...
Published: Sep 3, 2026 Updated: Sep 3, 2026
Google and Shopify announced the Universal Commerce Protocol (UCP) at the National Retail Federation’s Big Show on January 11, 2026, with Sundar Pichai using the keynote to introduce an open standard built to let AI agents discover, cart, and buy products from any participating merchant. Etsy, Wayfair, Target, and Walmart signed on as co-developers, and more than 20 additional partners, including Visa, Mastercard, Stripe, American Express, Best Buy, and The Home Depot, backed it at launch.
That timing wasn’t an accident. AI-referred traffic to US retail sites had grown 4,700% year over year by July 2025, according to Adobe Analytics. Consumers were already researching products through AI. UCP is Google and Shopify’s answer to what happens once that traffic actually needs to buy something.
Eight months on, UCP has added cart management, real time catalog queries, buy now pay later, a cross retailer Universal Cart, and a governing Tech Council that now includes Amazon, Meta, and Microsoft. It’s live today inside Google AI Mode and the Gemini app, and Shopify opened it to every developer in June. It’s also one piece of a broader shift toward agentic commerce, so if you want the wider picture before narrowing in on UCP specifically, that’s worth reading first. Here’s what UCP actually is, how it works, what’s changed since launch, and what it means for the product data behind your catalog.
UCP (Universal Commerce Protocol) is an open-source standard that defines how AI agents interact with merchants across the shopping journey, from product discovery through checkout and post-purchase support. Google and Shopify co-developed it, with Etsy, Wayfair, Target, and Walmart as founding partners, and it’s published on GitHub under the Apache 2.0 license, free for any developer to build on.
The problem UCP solves is fragmentation. Without a shared standard, every merchant needs a custom integration for every AI surface: Google, Gemini, Microsoft Copilot, and whatever comes next. UCP replaces that with a single, machine-readable capability profile that a merchant publishes once. Any UCP-compliant agent can read that profile and interact with the merchant’s catalog and checkout accordingly. One standard, any agent, any merchant, no bespoke integration work per platform.
When a shopper asks Gemini or Google’s AI Mode to find and buy something, the AI is acting as that shopper’s agent. UCP defines four stages for that interaction:
Merchants can expose these through a REST API, an MCP server, or Google’s Agent2Agent protocol, whichever fits their existing stack.
UCP launched with checkout as its primary capability. Since January, it has moved fast:
UCP is one of the fastest-moving specs in ecommerce right now. Anyone publishing on it, including this page, should expect to revisit it every few months.
UCP isn’t the only protocol built for AI-driven commerce. Like UCP, ACP or Agentic Commerce Protocol is a specification, not a product: it defines how an agent browses a merchant’s product feed, builds a cart, and completes checkout by passing a securely scoped payment token between buyer, agent, and merchant, with the merchant remaining the merchant of record throughout. ACP debuted alongside Instant Checkout, a feature that let ChatGPT users buy directly inside a conversation.
Here’s the part a lot of “UCP vs ACP” content published this year gets wrong: OpenAI retired Instant Checkout in March 2026, about six months after launch, after only a small number of Shopify merchants ever went live on it. ChatGPT now points shoppers toward product discovery and routes purchases to dedicated retailer apps inside the chat, Walmart’s among them, rather than completing checkout natively. ACP as a specification hasn’t disappeared; Stripe continues to build against it, and PayPal added support in 2025. But the checkout experience most comparison articles still describe as ACP’s flagship feature isn’t currently running.
UCP, by contrast, has kept its checkout live since January and kept adding capability on top of it. That’s a meaningful, current difference between the two protocols, not a settled one. Whether OpenAI ships a checkout successor, and whether merchants who built for ACP end up building for UCP instead, is still an open question. MCP vs UCP: How AI’s Protocols Affect Your Product Data goes deeper on how UCP and MCP specifically fit together, and what that split means for your own product data.
| UCP | ACP | |
|---|---|---|
| Co-developed by | Google and Shopify | OpenAI and Stripe |
| Announced | January 11, 2026, at NRF | September 2025 |
| License | Apache 2.0 | Apache 2.0 |
| What it covers | Discovery, cart, checkout, post-purchase | Cart and product feed, checkout, payment delegation |
| Flagship checkout feature | UCP-powered checkout in Google AI Mode and Gemini | Instant Checkout in ChatGPT |
| Status of that feature today | Live; expanded five times since launch | Retired by OpenAI in March 2026 |
| Where AI-driven purchases happen now | Native checkout inside Google surfaces | Routed to dedicated retailer apps inside ChatGPT |
| Backing coalition | 20+ partners; 10-member Tech Council including Amazon, Meta, Microsoft, Salesforce, Stripe | Stripe infrastructure; PayPal added support in 2025 |
UCP’s founding partners, Google, Shopify, Etsy, Wayfair, Target, and Walmart, were joined at launch by more than 20 endorsing partners across retail and payments, including Visa, Mastercard, Stripe, American Express, Best Buy, The Home Depot, Macy’s, Flipkart, Zalando, and Adyen. The Tech Council that steers the protocol’s direction has since grown to ten members with the April 2026 addition of Amazon, Meta, Microsoft, Salesforce, and Stripe.
That breadth matters. A coalition spanning retailers, payment networks, cloud platforms, and, as of April, two of Google’s biggest AI competitors, is a different proposition than a single vendor’s proprietary checkout. It’s a reasonable signal that UCP specifically is where the industry is converging, even if the full commercial picture is still being written.
UCP agents evaluate products using structured data: attributes, descriptions, availability, pricing, compatibility. The Catalog capability added in March lets an agent query that data conversationally, asking something like “is this available in a large” or “does this fit the model I already own.” If your product data can’t answer those questions clearly, because attributes are missing, specs are inconsistent, or availability is stale, the agent either skips the product or gives the shopper a wrong answer.
That makes UCP readiness a product data question, not a protocol implementation question. Shopify or Google Merchant Center handles the technical integration. What you control is whether the data behind that integration is complete, structured, and current enough for an agent to read and act on with confidence. That’s the same discipline a well-run PIM applies to product data generally, extended to a new channel.
Pimberly was named a Shopify Certified Technology Partner in August 2026, after Shopify assessed the connector against its criteria for Plus merchants: support availability, security and privacy review, performance, current API adoption, and proven results with Plus customers. Fewer than 100 apps in the Shopify App Store hold that certification, and Pimberly is currently the only UK-headquartered PIM in the directory.
Shopify built the UCP integration itself; a merchant’s job is making sure the product record behind it holds up. Pimberly governs that record: complete attribute sets, consistent categorization, current pricing and availability, and images and video pushed to Shopify’s CDN, so what Shopify exposes to an AI agent through UCP is accurate rather than whatever an unmanaged product feed happens to contain. Pimberly’s AI features apply that same governance across a catalog, regardless of which protocol ends up mattering most next year.
Q: What is UCP?
A: UCP (Universal Commerce Protocol) is an open source standard, co-developed by Google and Shopify, that lets AI agents discover products, build carts, complete checkout, and manage post-purchase actions across any participating merchant.
Q: Who created UCP?
A: Google and Shopify co-developed UCP, with Etsy, Wayfair, Target, and Walmart as founding partners. Google announced it at the National Retail Federation’s Big Show on January 11, 2026.
Q: Is UCP the same as ACP?
A: No. Both are open standards for AI-driven commerce, but they’re not interchangeable. ACP, co-developed by OpenAI and Stripe, launched alongside ChatGPT’s Instant Checkout feature in September 2025; OpenAI retired that feature in March 2026. UCP has kept its checkout live and continued expanding since its January 2026 launch.
Q: What is the difference between UCP and MCP?
A: MCP (Model Context Protocol) governs how an AI agent reads and understands data, including product data. UCP governs what an agent can actually do with a merchant once it has that understanding, like building a cart or completing checkout. They’re designed to work together rather than compete.
Q: Is UCP live now?
A: Yes. UCP-powered checkout is live for eligible retailers in Google AI Mode and the Gemini app in the United States. Expansion to Canada, Australia, and the UK was announced in May 2026 and is rolling out in phases.
Q: Do I need to implement UCP for my ecommerce store?
A: If you sell on Shopify, UCP support is built into the platform, and Shopify made it self-serve for developers in June 2026. If you sell elsewhere, implementation depends on your commerce platform and Merchant Center setup. Either way, the product data behind that integration determines whether an agent can act on it.
Q: How does product data affect UCP performance?
A: UCP agents query product data to answer specific questions, availability, sizing, compatibility, pricing, in real time. Missing attributes, inconsistent categorization, or outdated availability cause an agent to skip a product or answer incorrectly, regardless of how well the technical integration is built.
UCP is not a future concept. It’s live, it’s backed by a coalition that now includes two of Google’s biggest AI rivals, and it has shipped new capability roughly every two months since January. Google and Shopify solved the protocol problem. What determines whether your products actually show up when an AI agent goes looking, in UCP, in whatever comes after it, is whether your product data is structured, complete, and current enough to answer the questions an agent asks. That’s not a UCP problem specifically. It’s a product data problem, and it’s the one Pimberly is built to solve.
