Google AI releases Universal Commerce Protocol (UCP): an open-source standard designed to power the next generation of agent commerce

Can AI shopping agents go beyond sending product links and actually complete end-to-end trusted purchases inside chat? Universal Commerce Protocol, or UCP, is Google’s new open standard for agentic commerce. It gives AI agents and merchant systems a common language so that shopping queries can go from product discovery to confirmed order without custom integration for every retailer and every surface.

What problem is the UCP solving??

Today, most AI shopping experiences stop at recommendations. The agent collects the links, you handle the stock checks, coupon codes and checkout flow on the individual sites. Google’s engineering team describes this as an N by N integration bottleneck. Each new interactive surface requires different work from each merchant and payment provider.

UCP reduces it to an abstraction. Platforms like Gemini or AI Mode in search once integrated with the protocol. Businesses expose their commerce behavior once behind the UCP. Payment service providers and credential providers are integrated at the payment level. A single protocol can support multiple verticals such as shopping, travel or services.

Roles and main building blocks

The UCP core concept document defines four primary actors:

• Platform, which is the agent or application that organizes the user’s journey. Examples include AI shopping assistants and search surfaces.
• Business, which is the trader or service provider and the merchant of record.
• Credential provider, which manages personal data such as payment instruments and addresses.
• Payment service provider, which processes authorization, capture, and settlement.

On top of these roles, UCP offers three fundamental constructs:

• Capabilities, such as checkout, identity linking and ordering.
• Extensions, such as exemptions, fulfillment, or AP2 mandates, that expand a capability extends Field.
• Services that connect capabilities such as REST API, Model Context Protocol, and Agent2Agent to transports.

GitHub lists four initial key capabilities for repo purchases:

• Checkout manages checkout sessions, cart contents, totals and taxes.
• Identity linking uses OAuth 2.0 to allow agents to act on behalf of users.
• The order emits lifecycle events for shipment, return, and refund.
• Payment Token Exchange coordinates token and credential exchange between payment service providers and credential providers.

Commerce Lifecycle for AI Agent

Google Reference Implementation And the sample stores use a simple store to illustrate the UCP flow.

A typical agented checkout looks like this:

1. Agent receives business profile /.well-known/ucphe realizes dev.ucp.shopping.checkout and related extensions are available, and resolves the schema for those capabilities.
2. If the user has linked an account, the agent performs identity linking with an OAuth 2.0 scope that allows checkout and order read operations for that merchant.
3. The agent calls the checkout capability using REST or MCP bindings, passing line items, buyer fields, and any necessary context. The server returns a checkout object with line items, total, and candidate fulfillment options.
4. The agent applies discounts or loyalty benefits by implementing extensions that modify the compiled checkout schema, then asking the user to confirm the final order.
5. Payments are made through a payment handler that understands a specific payment instrument schema, such as a tokenized card. Once the payment service provider authorizes the transaction, the business creates an order.
6. The order capability emits webhook events for shipment and post-purchase adjustments, which the agent can present in the conversation.

From the user’s perspective, this puts the entire flow into a single conversation with clear consent steps.

Transportation, Payment and Security

The specification defines a transport layer with bindings to REST, the Model Context Protocol, Agent2Agent, and an embedded protocol that allows deeply customized merchant checkout experiences using UCP data structures.

For payments, the UCP agent integrates with the payment protocol. Payment architecture separates payment instruments from payment handlers, and uses mandates that fall within the scope of a specific checkout hash. This design supports proof of binding and reduces token replay risk, which is important when agents execute payments without the user directly in the browser.

Credential providers issue tokens and hold sensitive card data or addresses. Payment service providers consume those tokens and talk to the card network. UCP keeps these roles clear and uses verifiable credentials and signatures so that both agents and businesses have cryptographic evidence that they have been authorized.

key takeaways

• UCP is an open standard and open source specification from Google that defines a common commerce language for AI agents, businesses, payment providers, and credential providers across the entire purchase journey.
• The protocol is developed in collaboration with partners like Shopify, Etsy, Wayfair, Target, and Walmart, and is already supported by more than 20 ecosystem players including Visa, MasterCard, Stripe, PayPal, Best Buy, The Home Depot, Macy’s, and Zalando.
• The UCP model drives commerce through searchable capabilities such as checkout, identity linking and ordering, as well as extensible modules for discounts, fulfillment and membership, in which merchants and agents interact over a dynamically shared capability set through hosted profiles. .well-known/ucp.
• Protocol transport agnostic, it supports REST, JSON RPC, Model Context Protocol, and Agent2Agent, so the same capability schema can be reused across backend services, MCP tool calls into LLM agents, and can be used across agent-to-agent networks without rewriting business logic.
• Payment agents in UCP integrate with payment protocols and a modular payment handler design, which separates payment instruments from handlers and uses cryptographically proven mandates, so agents can autonomously execute checkout flows while preserving clear evidence of user consent.

check it out GitHub repo And technical details. Also, feel free to follow us Twitter And don’t forget to join us 100k+ ml subreddit and subscribe our newsletter. wait! Are you on Telegram? Now you can also connect with us on Telegram.

Check out our latest releases ai2025.devA 2025-focused analytics platform that models launches, benchmarks and transforms ecosystem activity into a structured dataset that you can filter, compare and export.

Asif Razzaq Marktechpost Media Inc. Is the CEO of. As a visionary entrepreneur and engineer, Asif is committed to harnessing the potential of Artificial Intelligence for social good. Their most recent endeavor is the launch of MarketTechPost, an Artificial Intelligence media platform, known for its in-depth coverage of Machine Learning and Deep Learning news that is technically robust and easily understood by a wide audience. The platform boasts of over 2 million monthly views, which shows its popularity among the audience.