What is the Agentic Commerce Protocol (ACP)? A Marketer’s Guide

Shopping is changing. Instead of browsing websites, comparing products, and checking out manually, consumers can increasingly rely on AI agents to discover products, make decisions, and even complete purchases. But for that to work, AI agents need a way to communicate with businesses and payment systems.

That’s where the Agentic Commerce Protocol (ACP) comes in. As an open standard for programmatic commerce, ACP helps connect buyers, AI agents, businesses, and payment providers. For marketers, this signals a shift from AI-powered product discovery to AI-assisted transactions. In this guide, we’ll break down what ACP is, how it works, and what it means for the future of marketing—including where AI connectors can fit into the broader agentic marketing stack.

Quick Answer: What is the Agentic Commerce Protocol?

AI is already changing how people shop, but most AI shopping experiences still stop at recommendations. An AI assistant might help someone find the right product, but the shopper still has to leave the conversation, visit a website, add the product to their cart, and complete the checkout themselves.

That’s the gap Agentic Commerce Protocol (ACP) is designed to address. Instead of simply helping shoppers decide what to buy, agentic commerce allows AI agents to participate in the transaction itself.

Traditional AI shopping:

AI recommends → user visits website → user checks out

Agentic commerce:

User expresses intent → AI discovers products → AI initiates checkout → transaction is completed

ACP provides standardized interfaces for agent-to-merchant interactions. Rather than requiring every AI agent and merchant to build a separate, custom integration, the protocol gives them a common way to communicate and handle commerce workflows. This makes it easier for AI agents to work with different merchants while allowing businesses to maintain control over their products, inventory, and fulfillment.

How Does ACP Work?

Think of ACP as the bridge between an AI agent and a merchant during a purchase. Instead of sending a shopper back and forth between different websites and checkout pages, the protocol gives the agent and merchant a standardized way to handle the buying process.

For a technical overview of how these parties interact, see the ACP architecture documentation: Documentation | Agentic Commerce Protocol 

1. Buyer expresses purchase intent

It starts with a simple request from the shopper, such as:

“Find me a large T-shirt under $40.”

The AI agent interprets what the buyer wants, including preferences such as product type, size, budget, or other requirements.

2. AI agent identifies products

The agent searches for relevant products and presents suitable options to the buyer. At this stage, the AI is helping with product discovery and narrowing down the choices.

3. Agent creates a checkout session

Once the buyer chooses a product, the agent can initiate a checkout session with the merchant. It can pass along relevant checkout information, such as the selected product, buyer details, shipping information, and other information needed to complete the purchase.

4. Merchant returns checkout information

The merchant responds with the details needed to complete the transaction, including:

• Products and quantities
• Pricing 
• Taxes
• Fulfillment or shipping options
• Available payment capabilities

This gives the AI agent the information it needs to present the actual purchase options to the buyer.

5. Buyer selects and confirms

The agent presents the available options, such as shipping methods and the final purchase amount. The buyer reviews the details and confirms that they want to proceed.

6. Payment is securely delegated

The payment can then be handled through delegated mechanisms, such as tokenized payment credentials, rather than requiring the AI agent to directly handle sensitive payment details. The important part is that payment information can be securely passed through the appropriate payment infrastructure while keeping the checkout experience within the agent-led flow.

7. Merchant completes and fulfills the order

Once the transaction is authorized, the merchant processes the order just as they would with a traditional purchase. The merchant remains responsible for the transaction and fulfillment, including preparing the product, shipping it, and handling the post-purchase experience. 

The Four Participants in ACP

ACP brings four key participants into the commerce flow. These roles are also outlined in the official ACP architecture documentation.

Each has a specific role, while the merchant continues to control the actual sale and fulfillment.

Participant	What they do
Buyer	Expresses purchase intent, reviews available options, and authorizes the purchase
AI Agent	Interprets the buyer’s intent, helps manage checkout, and communicates with the seller
Seller	Manages products, pricing, inventory, checkout, and order fulfillment
Payment Provider	Handles payment tokenization and secure payment processing

The Core Components of ACP

Here are the most important components that help AI agents and merchants handle the different stages of a transaction:

Agentic Checkout

Agentic Checkout manages the checkout lifecycle between the AI agent and the merchant. Instead of the agent simply sending the buyer to a website, it can interact with the merchant’s checkout system through a standardized flow.

The checkout lifecycle can include:

• Create checkout session: Start a new checkout for the selected product or products.
• Retrieve checkout: Get the latest details of an existing checkout.
• Update checkout: Modify details such as products, quantities, shipping information, or fulfillment options.
• Complete checkout: Confirm the purchase once the buyer has authorized it.
• Cancel checkout: End the checkout process when the buyer decides not to proceed.

This gives the AI agent a structured way to manage a purchase while the seller continues to control the underlying commerce experience.

The official ACP Checkout documentation describes this checkout lifecycle in more detail.

Delegated Payment

Payment is one of the more sensitive parts of agentic commerce, so ACP is designed to avoid giving AI agents unrestricted access to a buyer’s payment credentials.

Instead, payment can be handled through controlled payment tokens that represent the buyer’s authorization. These tokens can include constraints around how and where they can be used, helping limit the agent’s ability to make unauthorized transactions.

ACP’s Delegate Payment documentation explains how payment credentials can be tokenized with usage constraints.

The idea is simple: the AI agent can facilitate payment without needing to directly handle or store the buyer’s sensitive payment credentials.

Checkout Webhooks

The purchase doesn’t end when checkout is completed. Merchants also need a way to communicate what happens afterwards back to the AI platform.

ACP’s checkout webhooks allow merchants to send order lifecycle updates, such as:

• Order creation: Confirm that an order has been created.
• Order updates: Share changes to the order.
• Fulfillment status: Communicate whether an order has been processed, shipped, delivered, or otherwise updated.
• Refunds: Notify the AI platform when a refund has been issued.

This keeps the AI agent informed after the initial transaction, making it possible to support a more complete agent-led shopping experience from checkout through fulfillment and post-purchase updates.

The official Checkout Webhooks documentation covers these order lifecycle events.

ACP’s Key Features

Here are some of the features that make the protocol useful for both AI platforms and businesses.

Open and Interoperable

ACP is an open protocol designed as a common standard rather than a proprietary commerce system. This means AI agents and merchants can use the same framework to communicate instead of building a separate integration for every platform they want to work with.

The goal is to make agentic commerce more interoperable, so different AI systems and commerce businesses can work together without being locked into one ecosystem.

Merchant-Controlled

Even when an AI agent helps facilitate the purchase, the merchant remains the merchant of record. Businesses continue to control the parts of commerce they already manage, including:

• Products
• Pricing
• Inventory
• Fulfillment
• Customer relationships

This means ACP doesn’t replace a merchant’s existing commerce infrastructure. Instead, it gives AI agents a standardized way to interact with it.

Secure Payment Handling

Payment security is especially important when an AI agent is involved in a transaction. ACP supports delegated payment flows that use controlled payment credentials or tokens and authorization constraints, allowing AI agents to facilitate payments without gaining unrestricted access to the buyer’s underlying payment credentials.

These controls can limit how a payment credential or token is used, helping keep sensitive payment information protected while still allowing the transaction to move forward.

Capability Negotiation

Not every merchant or payment setup supports the same checkout features. ACP allows agents and sellers to communicate their supported capabilities before proceeding with a transaction.

For example, they can indicate whether they support:

• 3DS authentication
• Biometric authentication
• Address verification
• Specific payment methods
• Additional checkout extensions 

In simple terms, it’s like the agent and merchant checking what each side can handle before completing the purchase. This helps the checkout flow adapt to the capabilities available rather than assuming every seller supports the same features.

The ACP documentation also describes how supported interventions and extensions can be communicated between parties.

Extensible

Commerce doesn’t stop at the basic checkout flow. Businesses may also want to support features such as discounts, loyalty programs, and other specialized commerce experiences.

ACP is designed to be extensible, allowing additional capabilities to be added through extensions without changing the core protocol. This gives merchants and platforms room to build more advanced agentic commerce experiences as the ecosystem develops.

Why Does ACP Matter for Marketers?

For marketers, ACP isn’t just another technical protocol to keep an eye on. It points to a potential shift in where product discovery and purchasing happen.

AI Could Become a New Commerce Discovery Channel

Instead of searching Google, browsing a marketplace, and comparing dozens of product pages, a customer could simply tell an AI agent what they’re looking for. The agent could understand the request, find relevant products, compare options, and help complete the purchase.

That means brands may eventually need to think beyond where their products rank and consider how easily AI agents can understand and recommend them. 

The Customer Journey Becomes Conversational

Traditional ecommerce generally follows a fairly predictable path:

Search → Product page → Cart → Checkout → Purchase

Agentic commerce introduces a more conversational journey:

Prompt → AI recommendation → Agentic checkout → Purchase

The customer may not need to visit multiple websites or manually move between product pages and checkout screens. Instead, much of the journey can happen through an AI interface.

For marketers, this could change how they think about everything from product discovery and messaging to conversion optimization.

Product Information Becomes More Important

AI agents need reliable information to make accurate recommendations. A product may look great to a human browsing a website, but if key details are incomplete, outdated, or difficult for an agent to interpret, the product may not be recommended accurately.

Brands should therefore pay close attention to information such as:

• Product availability
• Pricing 
• Product attributes
• Fulfillment options
• Discounts
• Shipping and return policies

In other words, product data isn’t just something that helps a customer make a purchase. It could increasingly become something that helps an AI agent decide whether to recommend the product in the first place.

This also means marketers may need better ways to bring live product, search, advertising, and analytics data into the AI workflows they already use.

Conversion Friction Could Decrease

Every additional step between discovering a product and completing a purchase creates an opportunity for a customer to drop off.

If a shopper can go from discovering a product in an AI conversation to reviewing checkout details and confirming a purchase without leaving that interface, the buying journey could become considerably smoother.

For brands, that could mean fewer interruptions between intent and transaction. And as agentic commerce develops, reducing that friction could become an increasingly important part of the customer experience.

ACP vs Traditional Ecommerce

	Traditional Ecommerce	Agentic Commerce
Product discovery	Search engines and ecommerce websites	AI agents alongside traditional channels
Interaction	Website or app	Conversational AI interface
Checkout	Customer navigates the checkout flow	AI agent coordinates the checkout
Payment	Customer enters or authorizes payment	Delegated payment flow
Merchant	Controls the transaction	Remains the merchant of record

ACP vs MCP: What’s the Difference?

	MCP	ACP
Primary purpose	Connect AI applications to external tools and data	Standardized AI-driven commerce transactions
Main interaction	AI <—> tools/data	Buyer <—> AI agent <—> seller
Typical use	Retrieve information or perform actions through connected tools	Discover products, initiate checkout, handle payment, and complete purchases
Commerce role	Can provide access to commerce-related tools	Defines the commerce transaction flow
Example	AI accesses marketing analytics through an MCP connection	AI completes a purchase through an ACP-enabled merchant

For marketers, this distinction also explains where tools such as Smacient’s Claude Connector fit. MCP provides the underlying connection framework, while a Claude Connector can give Claude access to live data and external marketing tools through that framework.

That distinction is important for marketers because MCP can expand what an AI application can access, while ACP is specifically designed to help move an AI-led shopping experience from product discovery to completed purchase.

How Can MCP and ACP Work Together?

MCP and ACP aren’t competing standards, and ACP shouldn’t be described as being built on MCP. Instead, they can play different roles within the same AI-powered workflow.

ACP defines how the commerce transaction works from checkout through payment and order completion. MCP provides a standardized way for an AI application to interact with external tools and services.

For example, an AI application could use MCP to access a service that exposes commerce functionality, while ACP governs the actual agentic checkout and transaction flow. ACP’s documentation also supports publishing checkout configuration through REST or MCP, giving developers different ways to expose the commerce functionality to AI applications.

So, in simple terms:

MCP = how AI connects to tools and services

ACP = how AI-enabled commerce transactions are handled 

For marketers, this distinction becomes useful when looking at the broader agentic marketing stack. ACP can handle the commerce transaction layer, while MCP can connect AI applications to the marketing tools and live data they need before and after a purchase.

Smacient’s Claude Connector is one example of this MCP-based approach, connecting Claude with live marketing data from platforms such as Google Ads, Meta Ads, Google Analytics 4, Google Search Console, and Google Keyword Planner.

What Does ACP Mean for the Future of AI Marketing?

The next shift in the AI world could be even bigger: AI moving from helping marketers make decisions to actively participating in customer journeys and transactions. But for that broader agentic journey to work, AI also needs access to the data and tools marketers use every day.

From AI-Assisted Marketing to Agentic Marketing

The evolution can be thought of in three stages:

AI-assisted marketing

AI analyzes data and helps marketers make better decisions.

↓

AI-powered workflows

AI connects with external tools and can execute marketing tasks, such as pulling reports, updating campaigns, or analyzing customer data.

↓

Agentic commerce 

AI can participate directly in product discovery, checkout, payment, and purchasing on behalf of the buyer.

ACP fits into this third stage by providing a standardized way for AI agents and merchants to handle commerce transactions.

What Could Change for Marketers?

Product Discovery

Products may increasingly be discovered through AI conversations alongside search engines, marketplaces, and social platforms. Marketers may need to make product information easier for AI systems to understand and evaluate.

Performance Marketing

If AI agents increasingly influence purchase decisions, marketers may need to rethink how they measure visibility, attribution, and conversions across AI-led journeys. Having live advertising and analytics data available inside AI workflows could also make it easier to analyze performance alongside these emerging customer journeys.

Personalization

AI agents can understand a shopper’s specific intent and preferences during a conversation, creating opportunities for more contextual product recommendations.

Customer Acquisition

AI interfaces could become another channel through which customers discover and purchase from brands, potentially changing how businesses think about acquisition beyond traditional search and advertising.

Conversion Optimization

When an AI agent can take a customer from product discovery directly into checkout, there may be fewer steps between purchase intent and conversion. This could make reducing friction even more important.

Retail Media

As AI becomes part of product discovery, brands and retailers may eventually need to consider how sponsored products, recommendations, and advertising fit into agent-led shopping experiences.

Customer Experience

The shopping journey could become less about navigating websites and more about having a conversation. Instead of asking customers to find the right product themselves, brands may increasingly need to make sure AI agents can understand their products, policies, and offerings accurately.

Where Smacient’s Claude Connector Fits Into the Agentic Marketing Stack

Commerce is only one part of the agentic customer journey. Before a customer ever reaches checkout, marketers still need AI to work with the tools they use to understand audiences, run campaigns, measure performance, and optimize discovery.

That’s where Smacient’s Claude Connector fits into the broader agentic marketing stack. Using MCP, it connects Claude with live marketing data from platforms including:

• Google Ads for campaign and performance data
• Meta Ads for advertising insights and campaign management
• Google Analytics for website and customer behavior data
• Google Search Console for search performance and SEO insights
• Google Keyword Planner for search volume, trends, and keyword opportunities

You can also get started with 30 free credits every month, with no credit card required.

Instead of exporting reports or manually pasting data into Claude, marketers can ask questions in plain English and have Claude pull the relevant data through the connector. Smacient’s Claude Connector currently offers a wide range of tools that work with Claude.ai, Claude Desktop, Claude Code, and Claude Cowork.

For example, a marketer could use it to analyze live Google Ads performance, pull Google Search Console opportunities, research Google Keyword Planner data, or compare Meta Ads performance, all within Claude. 

This is where the distinction between ACP and MCP becomes useful: ACP focuses on the commerce transaction, while Smacient’s Claude Connector uses MCP to bring marketing data and tools into Claude. Together, they illustrate how different parts of the customer journey can become increasingly AI-driven.

ACP Use Cases for Businesses

Business Use Case	Potential Application
Ecommerce and retail	AI-assisted product discovery, comparison, checkout, and purchase
Digital products	Agent-assisted purchases of software, subscriptions, and other digital products
Food and delivery	Potential agent-led ordering and coordination of fulfillment
Travel and bookings	Potential agent-assisted reservation and purchase flows
Subscription commerce	Potential support for recurring products and services, including subscription purchases

What Businesses Need to Consider Before Adopting ACP

Here are a few questions businesses should consider:

Consideration	What to Ask
Commerce infrastructure	Is your existing commerce stack API-ready and able to support programmatic interactions?
Product and inventory data	Can AI agents reliably access accurate information about products, pricing, availability, and fulfillment?
Checkout	Can your checkout infrastructure support interactions initiated through an AI agent?
Authentication and payments	How will you handle authentication, authorization, and payment security without exposing sensitive credentials?
Agent access and control	How will you decide which AI agents can interact with your commerce systems and complete transactions?
Measurement	How will you identify, attribute, and measure conversions that originate from AI-driven shopping journeys?
AI and marketing data	Can your marketing and analytics systems provide the data needed to measure and optimize AI-driven customer journeys?

Businesses can also review the official ACP seller implementation guide to understand the infrastructure and endpoints involved.

Decision Framework

ACP makes sense if…	ACP may not make sense if…
You sell products or services online	Your business doesn’t involve online transactions
Your products, pricing, and inventory data are available digitally	Product and inventory information is mostly manual or unstructured
Your checkout can support API-based interactions	Your checkout isn’t accessible programmatically
You want to explore AI-led product discovery and purchasing	AI-driven commerce isn’t relevant to your customer journey yet
You’re investing in AI-powered customer experiences	Your current priority is improving traditional ecommerce channels
You have the infrastructure to support secure payments and order fulfillment	Your payment, authentication, or fulfillment systems aren’t ready for automated workflows
You want to experiment with AI-driven product discovery and acquisition channels	You don’t yet have reliable product data or ecommerce fundamentals in place

AI interfaces could become more than places to ask questions. They could become new destinations for discovering and buying products. As conversational product discovery evolves, shoppers may be able to move from intent to recommendation to checkout with fewer steps.

Protocols like ACP could help standardize how AI agents and businesses interact, while making structured product, pricing, inventory, and commerce data increasingly important. For marketers, this could also create a new AI-driven acquisition channel to consider alongside search, social, marketplaces, and traditional advertising.

Agentic commerce is still evolving, so it’s too early to say exactly how the ecosystem will develop. But one thing is clear: the way customers discover and purchase products is becoming increasingly shaped by AI.

For marketers, this also means preparing the data and AI workflows that support customers before, during, and after the transaction.

Want to explore how Claude can work with live marketing and commerce data? Check out more Smacient guides:

• Best Tools to Extract Quick Commerce Data Without an API
• Best AI tools for Ecommerce
• Best AI Agents for E-Commerce

FAQs

Disclosure – This post contains some sponsored links and some affiliate links, and we may earn a commission when you click on the links, at no additional cost to you.