Google Expands Universal Commerce Protocol With Cart, Catalog, and Identity Linking as Agentic Shopping Takes Shape
Google added Cart, Catalog, and Identity Linking capabilities to its Universal Commerce Protocol, expanding the open standard that lets AI agents complete purchases across retailer platforms.
Google on March 19 released three new optional capabilities for the Universal Commerce Protocol, the open standard for AI-driven shopping that the company launched in January alongside Shopify, Target, Walmart, and more than 20 other retailers and payment processors. The additions — Cart, Catalog, and Identity Linking — mark the protocol’s first major functional expansion and bring it closer to enabling AI agents to handle multi-step purchase workflows without human intervention.
What the New Capabilities Do
The Cart capability allows AI agents to save or add multiple items to a shopping basket from a single store in one operation, mirroring how human shoppers typically build orders rather than checking out one product at a time. Catalog enables agents to retrieve real-time product details including variants, pricing, and inventory directly from retailer systems, eliminating reliance on stale product feeds. Identity Linking lets shoppers carry loyalty memberships and rewards across UCP-integrated platforms, so that an AI agent completing a purchase can automatically apply a shopper’s existing benefits.
Google also announced a simplified onboarding process in Merchant Center that will roll out over the coming months, and confirmed that platform partners Salesforce, Stripe, and Commerce Inc. will implement UCP support.
The Protocol’s Architecture
UCP defines a set of standardized primitives — checkout sessions, line items, discount application, fulfillment tracking, and order management — that allow any AI agent to interact with any compliant merchant backend through a common language and functional interface. Merchants publish capability profiles at a well-known endpoint, and agents discover those profiles to negotiate which features both sides support.
The protocol separates payment instruments from payment handlers, allowing integration with diverse processors. Google’s reference implementation currently powers native checkout within AI Mode in Google Search and the Gemini app using Google Pay, with PayPal integration planned for future releases. Retailers remain the seller of record in all transactions.
A distinctive design choice is UCP’s transport flexibility. The specification supports REST APIs, Google’s Agent2Agent protocol, and Anthropic’s Model Context Protocol as communication bindings, meaning agents built on different frameworks can interact with the same merchant endpoints. Security is enforced through tokenized payments, verifiable credentials, and cryptographic request signing.
Scale and Industry Support
The protocol was announced on January 11, 2026, at the National Retail Federation conference and co-developed with Shopify, Etsy, Wayfair, Target, and Walmart. More than 20 additional organizations endorsed it at launch, including Adyen, American Express, Best Buy, Flipkart, Mastercard, Stripe, The Home Depot, Visa, and Zalando.
Shopify’s engineering team contributed foundational checkout architecture drawn from the company’s Checkout Kit, designing a layered capability model that allows independent versioning and extension composition without breaking backward compatibility. Merchants on Shopify’s platform can expose their storefronts to UCP-compliant agents through capability profiles that declare supported features, with the system negotiating available payment options per transaction based on cart contents and buyer location.
Google’s Shopping Graph, which underpins the protocol’s product discovery layer, now contains more than 50 billion product listings with over 2 billion refreshed every hour. The company processed 8.3 trillion tokens through its commerce systems in December 2024 and reported that figure had grown to more than 90 trillion one year later.
From Pages to Protocols
The UCP expansion reflects a broader shift in how online commerce infrastructure is being rebuilt for an era of AI-mediated transactions. Traditional e-commerce relies on human shoppers navigating web pages, comparing products visually, and completing checkout forms. Agentic commerce replaces that browsing model with structured, machine-readable interactions where AI agents handle discovery, comparison, and purchase on behalf of users.
Google AI Lead James Massey has stated that UCP allows smaller brands to become discoverable to AI agents without massive ad budgets, addressing what the company calls the N-by-N integration problem: the requirement for separate connections between each platform and each sales channel that has historically limited agentic shopping deployment.
Checkout sessions in the protocol progress through defined states — incomplete, requires escalation, and ready for complete — allowing agents to resolve issues via API or hand off to human buyers through a continuation URL when they encounter situations beyond their capabilities. This escalation mechanism is intended to maintain consumer trust while enabling autonomous transaction completion in straightforward cases.
What Remains Ahead
UCP checkout on Google’s surfaces remains in early access, limited to select U.S. merchants who have completed an eligibility review through Merchant Center. The specification itself is published at ucp.dev and the GitHub repository is open for community contributions. Google has indicated that the program will expand globally in the coming months and add capabilities for discovering related products within the checkout experience, applying loyalty rewards, and powering custom shopping experiences.
The protocol’s open-source governance and multi-vendor backing distinguish it from proprietary commerce APIs, but its long-term adoption will depend on whether competing AI platforms — including those from Amazon, Meta, and Apple — choose to adopt UCP or develop rival standards. For now, Google’s March 19 update moves the protocol from an abstract specification toward a functional commerce layer that AI agents can use to shop on behalf of the humans who deploy them.