Facebook Community
Change Log
Help Center
Agentic Commerce Protocol (ACP): The New Rules Behind AI Shopping
AI can help shoppers find products, compare options, and make buying decisions without visiting multiple websites. But what happens when a customer decides to purchase something directly through an AI assistant? The AI needs a reliable way to communicate with the merchant, verify product information, and support a transaction without taking control away from the buyer.
That's where the Agentic Commerce Protocol (ACP) comes in. Developed by OpenAI and Stripe, ACP introduces a common standard for connecting AI shopping experiences with existing ecommerce systems. As AI becomes a new shopping channel, understanding how these connections work is increasingly relevant for online businesses.
In this guide, we'll explore how ACP works, the security mechanisms behind AI-assisted transactions, how it compares with UCP and MCP, and what ecommerce merchants should know before adopting it.
Key Takeaways
- ACP connects AI shopping with ecommerce systems: It standardizes supported interactions involving product information, checkout, payments, and order updates.
- Merchants remain in control: AI can facilitate purchases, but buyer authorization, payment processing, and order fulfillment remain essential parts of the transaction.
- ACP, UCP, and MCP serve different purposes: ACP and UCP address commerce interoperability, while MCP connects AI agents with external tools and data.
- Adoption requires preparation: Accurate product data, compatible integrations, and platform eligibility determine how merchants can participate in AI-powered commerce.
What Is the Agentic Commerce Protocol (ACP)?
The Agentic Commerce Protocol (ACP) is an open standard developed by OpenAI and Stripe in 2025 that enables AI shopping platforms to communicate with merchants' existing ecommerce systems. It establishes a shared set of rules for exchanging product information and supporting transactions without requiring merchants to replace their checkout or payment infrastructure.
A simplified illustration of the Agentic Commerce Protocol (ACP) showing how an AI agent connects shopper requests with product data, merchant systems, and checkout actions.
In simple terms, ACP helps AI assistants and online stores exchange the information needed to support a purchase. Its core capabilities include:
- Product discovery: Merchants provide structured product data, including descriptions, prices, variants, and availability, to supported AI shopping platforms.
- Checkout integration: AI platforms can create and update checkout sessions using information verified by the merchant.
- Delegated payments: Supported payment mechanisms allow transactions to proceed within authorized limits.
- Order management: Merchants can share order confirmations and status updates with AI shopping platforms.
For example, when a shopper asks ChatGPT to purchase a pair of running shoes, ACP can support the exchange of checkout information between the AI platform and the retailer. The merchant still controls pricing, payment processing, and order fulfillment, while the shopper authorizes the purchase.
ACP is not an AI shopping assistant or a payment gateway. It is the technical standard that helps these systems work together.
If you want the wider picture first — the types of AI shopping agents, real examples in 2026, and how to get a store ready — start with our overview of agentic commerce, then come back here for the protocol mechanics.
Explore more: New to AI-powered shopping? Read our guide to agentic shopping to understand how AI agents are changing the customer buying journey.
How Does the Agentic Commerce Protocol Work?
The Agentic Commerce Protocol (ACP) connects AI-powered shopping experiences with the ecommerce systems that manage products, payments, and orders. Rather than replacing a merchant's existing infrastructure, it provides standardized ways for supported AI platforms to exchange information and coordinate purchases.
To understand how the Agentic Commerce Protocol works, consider a shopper looking for running shoes under $150 through an AI assistant. The process involves four main participants: the shopper, the AI shopping platform, the merchant, and the payment service provider (PSP).
The Agentic Commerce Protocol workflow illustrates how AI shopping moves from product discovery and comparison to payment authorization and order completion, using merchant data and checkout systems.
Step 1. The Shopper Searches for a Product Through AI
The shopping journey begins when a customer describes what they want in natural language. For example: "Find me a pair of running shoes under $150 for everyday training."
The AI assistant interprets the request and identifies relevant requirements, including the customer's budget, intended use, and product preferences. It can then present suitable options, explain their differences, and ask follow-up questions when necessary.
Unlike a traditional keyword search, this interaction allows customers to refine their choices through conversation.
However, understanding the request and recommending products are primarily functions of the AI shopping platform. ACP supports the commerce integrations behind the experience rather than performing the AI's reasoning itself.
Step 2. The AI Platform Uses Structured Product Data
To recommend products accurately, the AI platform needs reliable information from participating merchants.
Under ACP's product feed specifications, merchants can provide structured catalog data containing:
- Product identifiers and titles: Information that identifies each product.
- Descriptions and attributes: Features, materials, sizes, and other relevant specifications.
- Pricing and availability: Current prices, inventory status, and available variants.
- Product images and URLs: Visual references and links to product listings.
- Fulfillment information: Supported delivery and shipping details.
For example, the AI could use this information to find running shoes that match the shopper's budget and preferred size.
Product feeds are generally supplied and refreshed through an integration process, rather than retrieved from scratch for every conversation. Keeping that information accurate is important because recommendations based on outdated prices or unavailable products can create problems later in checkout.
Step 3. ACP Creates and Updates a Checkout Session
Once the shopper selects a product, the AI shopping platform can initiate a checkout session with the merchant through the Agentic Checkout Specification.
A checkout session represents the current state of a purchase, including the selected items and information needed to complete the order.
The exchange typically involves:
- Creating a session: The AI platform sends the selected products and quantities to the merchant's checkout endpoint.
- Receiving checkout details: The merchant returns the current items, prices, discounts, taxes, shipping options, and total.
- Updating the session: If the shopper changes the quantity, delivery address, or shipping method, the platform requests an updated checkout state.
Suppose the selected running shoes cost $120. After the shopper provides a delivery address, the merchant calculates applicable shipping charges and taxes. The AI interface can then display the updated total before the customer confirms the purchase.
Importantly, the merchant remains the authoritative source for checkout information. The AI platform should not invent final prices or assume that previously displayed inventory information is still accurate.
Step 4. The Shopper Reviews and Authorizes Payment
After the checkout details are confirmed, the shopper can review the order and authorize payment through the supported checkout experience.
ACP includes a Delegated Payment Specification designed to let payment information move securely between participating systems while maintaining appropriate restrictions.
In a supported delegated payment flow:
- The shopper reviews the purchase and authorizes the transaction.
- A delegated payment request specifies restrictions, such as a maximum chargeable amount and expiration time.
- A compatible payment service provider handles the payment credentials and produces a payment token.
- The merchant uses its existing payment infrastructure to process the authorized payment.
This means the AI platform helps coordinate the purchase, while the merchant and its payment provider handle the actual financial transaction. The AI agent does not become the payment processor or gain unrestricted permission to spend the shopper's money.
Step 5. The Merchant Confirms the Order and Provides Updates
Once payment and checkout are completed, the merchant determines whether the order can be accepted and returns the result to the AI shopping platform.
The shopper can then receive an order confirmation or an explanation if the transaction was declined.
The merchant remains responsible for subsequent activities, including:
- Processing and fulfilling the order.
- Managing shipping and delivery.
- Handling cancellations, refunds, and returns.
- Providing customer service.
- Sharing supported order status updates.
ACP's order events and webhooks allow the merchant's systems to communicate relevant order changes to the AI shopping platform. This helps keep the information shown to shoppers consistent with the merchant's order records.
The responsibilities across the shopping journey can be summarized as follows:
| Stage | AI Shopping Platform | Merchant |
|---|---|---|
| Product discovery | Interprets requests and recommends products | Provides structured product information |
| Checkout | Creates and updates checkout sessions | Verifies prices, availability, and order details |
| Payment | Facilitates the authorized payment flow | Processes payment through its PSP |
| Order management | Displays confirmations and supported updates | Manages fulfillment and post-purchase activities |
The key distinction is that ACP standardizes supported interactions between AI shopping platforms and merchant systems, while each party retains its own responsibilities. The AI can help customers move through the shopping journey, but the merchant remains responsible for accepting and fulfilling purchases.
That division of responsibility also makes security an essential part of the protocol, particularly when AI systems handle checkout information and payment authorization.
Learn more: How To Integrate AI into a Website To Boost Performance and Sales
How Does ACP Keep AI Shopping Transactions Secure?
When shoppers purchase products through an AI assistant, security involves more than protecting payment details. Merchants must also verify that a transaction is authorized, confirm that the order information is accurate, and prevent duplicate or unauthorized requests.
The Agentic Commerce Protocol (ACP) addresses these concerns through payment restrictions, merchant-controlled transactions, and secure communication standards.
ACP helps protect AI shopping transactions through buyer authorization, merchant payment controls, and secure communication between AI platforms and ecommerce systems.
Buyer Authorization and Payment Limits
ACP's Delegated Payment Specification helps ensure that payment credentials are used only within the limits of an authorized purchase.
In supported implementations, several safeguards apply:
- Buyer authorization: Shoppers review and approve their purchases before payment is completed.
- Single-use payment tokens: Tokens are designed for a specific transaction rather than unrestricted future purchases.
- Spending limits: Payment requests include a maximum chargeable amount.
- Expiration limits: Payment authorization is restricted to a defined period.
Merchant Control Over Payments and Orders
ACP does not transfer responsibility for transactions to the AI shopping platform. The merchant remains the Merchant of Record, meaning it continues to manage the commercial and payment responsibilities associated with each order.
These responsibilities include:
- Validating order details and payment information.
- Running fraud and risk checks.
- Accepting or declining transactions.
- Processing refunds, chargebacks, and fulfillment.
Secure Communication and Transaction Validation
ACP's checkout specification also defines technical safeguards for communication between AI platforms and merchant systems:
- Authentication: Verifies that API requests come from authorized systems.
- Signature verification: Helps detect requests that have been altered.
- Input validation: Checks that submitted transaction information meets the required format and rules.
- Idempotency: Prevents repeated requests from unintentionally creating duplicate transactions.
- Secure connections: Checkout endpoints use HTTPS to protect data in transit.
These mechanisms provide a foundation for safer AI-assisted commerce, although their effectiveness still depends on correct implementation by the AI platform, merchant, and payment provider.
As AI shopping develops, ACP is also becoming part of a broader ecosystem of standards. Understanding how it compares with UCP and MCP helps clarify which protocol handles transactions and which connects AI agents to other ecommerce tools.
ACP vs UCP vs MCP: What's the Difference?
As AI shopping evolves, several protocols are emerging to help AI agents interact with external systems. Three important standards are the Agentic Commerce Protocol (ACP), Universal Commerce Protocol (UCP), and Model Context Protocol (MCP).
Although they share a goal of making AI systems more useful, they serve different purposes. ACP and UCP focus on commerce interactions, while MCP provides a broader way for AI agents to connect with external tools and data.
ACP vs UCP: Two Approaches to AI Commerce
Both ACP and UCP aim to make shopping through AI assistants possible without requiring merchants to rebuild their existing ecommerce infrastructure. However, their specifications differ in scope, architecture, and implementation.
| Feature | ACP (Agentic Commerce Protocol) | UCP (Universal Commerce Protocol) |
|---|---|---|
| Developed by | OpenAI and Stripe | Google and Shopify, with industry partners |
| Primary focus | Connecting AI shopping experiences with merchant commerce systems | Standardizing commerce interactions across AI platforms and businesses |
| Architecture | Product data integration and agentic checkout specifications | Modular capabilities, services, and extensions |
| Checkout | Standardized checkout sessions and order updates | Checkout capabilities with support for additional commerce workflows |
| Payments | Delegated payments through compatible payment providers | Payment handlers and tokenized payment integrations |
| Example ecosystem | ChatGPT shopping integrations | Google Search AI Mode and Gemini |
Sources: OpenAI ACP documentation and Google UCP documentation.
The main difference lies in their design approaches:
- ACP defines supported interactions between AI shopping platforms and merchant systems, including product feeds, checkout sessions, and payment authorization.
- UCP uses a modular architecture, allowing businesses to implement supported commerce capabilities and extensions through different communication methods.
Neither protocol automatically makes every online store compatible with every AI assistant. Availability depends on platform support, merchant eligibility, and the capabilities implemented.
Where Does Model Context Protocol (MCP) Fit In?
Unlike ACP and UCP, Model Context Protocol (MCP) is not specifically designed for ecommerce transactions. It is an open standard that allows AI agents to connect with external applications, retrieve information, and perform supported actions.
For example, an AI agent could use MCP to access product data, conduct research, or work with a connected ecommerce tool. MCP defines how that tool connection works, rather than providing a dedicated commerce checkout standard.
A practical example is GemPages MCP, which connects compatible AI agents with Shopify and GemPages to support ecommerce content and page creation. GemPages MCP is currently in beta, and access is being introduced gradually.
GemPages MCP connects AI agents with Shopify and GemPages to support content research, HTML generation, and editable landing page creation.
It offers three core capabilities:
- Content Research: Analyze product information, competitors, customer reviews, pain points, and buyer motivations to develop content ideas.
- Content to HTML: Transform completed advertorials, listicles, or comparison articles into structured HTML landing pages, using a template that matches the content type.
- AI Page Generation: Create editable GemPages pages using HTML, images, or webpage URLs as references. Only use webpages and assets that you own or have permission to reuse.
For instance, a Shopify merchant could use GemPages MCP to research customer objections, turn the findings into an advertorial, and create an editable landing page. These activities support the merchant's content production process but do not enable ACP checkout.
Importantly, these protocols are not necessarily competing alternatives. UCP, for example, supports MCP as a communication method alongside APIs. This means MCP can help connect systems while commerce-specific standards define the transactions those systems support.
For ecommerce merchants, understanding these distinctions helps determine which technologies are relevant to their goals, whether that's enabling AI-assisted purchases, connecting business tools to AI agents, or improving content production. The next consideration is what businesses can realistically implement today and how to prepare their stores for agentic commerce.
Explore more: How to Use Claude Code with GemPages to Build Smarter Shopify Pages & Real Use Cases
What Does ACP Mean for Ecommerce Merchants?
The Agentic Commerce Protocol creates new opportunities for merchants to reach customers who increasingly use AI assistants to research and purchase products. Instead of relying entirely on shoppers visiting their websites, businesses can participate in supported AI shopping experiences while continuing to use their existing commerce infrastructure.
However, adopting ACP involves more than connecting a store to an AI platform. Merchants need to understand its current capabilities, integration requirements, and limitations before investing in implementation.
Opportunities and Current Limitations of ACP
ACP offers several potential advantages for ecommerce businesses:
- A new shopping channel: Merchants can make their products available through supported AI shopping experiences, reaching customers who prefer conversational product discovery.
- A shorter path to purchase: Supported checkout integrations can allow shoppers to move from product selection to payment without leaving the AI experience.
- Greater control over transactions: Merchants retain responsibility for pricing, inventory validation, payment processing, and order fulfillment.
- Standardized integrations: ACP provides common specifications for supported commerce interactions, potentially reducing the need for entirely custom integrations.
However, merchants should distinguish between what ACP makes technically possible and what is currently available.
According to OpenAI's ACP documentation, building with ACP is open to everyone, but Instant Checkout in ChatGPT is currently limited to approved partners. Product feed onboarding also has its own approval requirements.
Additionally, ACP does not guarantee that products will appear in AI recommendations or generate more sales. Product visibility depends on the AI platform's discovery systems, available catalog data, and other relevance factors.
For merchants, the opportunity is significant, but adoption should be based on actual platform support rather than the assumption that ACP automatically enables AI checkout for every online store.
How to Prepare Your Store for Agentic Commerce Protocols
Even if a business cannot immediately participate in ACP-powered checkout, it can prepare its ecommerce infrastructure for future integrations.
The following checklist covers the areas merchants should prioritize:
| Preparation Area | What Merchants Should Do |
|---|---|
| Product data | Maintain accurate titles, descriptions, product IDs, variants, prices, and inventory information. |
| Catalog integration | Review product feed specifications and establish processes for updating product information. |
| Checkout infrastructure | Evaluate whether existing systems can create, update, and complete checkout sessions through supported APIs. |
| Payment processing | Check payment provider compatibility and requirements for delegated payment authorization. |
| Order management | Ensure order confirmations, cancellations, fulfillment statuses, and refunds can be tracked reliably. |
| Security | Review API authentication, request validation, fraud prevention, and duplicate transaction safeguards. |
| Platform eligibility | Check onboarding requirements and supported integrations before committing development resources. |
Product information is a particularly useful starting point. OpenAI recommends accurate descriptions, consistent seller information, properly modeled product variants, and valid product URLs for high-quality catalog integrations.
Merchants can also improve the content customers encounter before purchasing. For example, Shopify businesses can use tools such as GemPages MCP to research customer needs and create editable landing pages with AI agents.
These content workflows complement broader agentic commerce preparation, but they are separate from ACP checkout integration.
Ultimately, merchants should prioritize accurate product data, reliable transaction systems, and compatibility with supported AI commerce platforms rather than treating ACP as a feature they can simply switch on.
Conclusion
The Agentic Commerce Protocol (ACP) helps connect AI-powered shopping experiences with existing ecommerce systems through standardized product data, checkout, payment, and order interactions. While ACP introduces new possibilities for AI-assisted purchases, merchants remain responsible for processing transactions and fulfilling orders.
As standards such as ACP, UCP, and MCP continue to develop, ecommerce businesses should focus on understanding their different roles and preparing the systems they already use. Accurate product information, secure checkout infrastructure, and reliable integrations will be essential as AI becomes a more established shopping channel. For more guides on preparing your store for AI-driven commerce, browse the GemPages blog.
FAQs about the Agentic Commerce Protocol
