From discovery to checkout: Understanding the Agentic Commerce Protocol (ACP)

As consumers increasingly turn to AI-driven product discovery and shopping — and as agentic commerce begins to take shape— connecting brands, commerce platforms and AI channels in a standardized way is crucial. 

The Agentic Commerce Protocol, or ACP for short, is an open standard launched by OpenAI and Stripe (with commercetools as a launch partner) that helps businesses transact with any AI agent or payment processor via ChatGPT and Microsoft CoPilot. 

The goals are simple: For businesses, ACP streamlines payments for single-item purchases in generative AI environments by leveraging a delegated payment model. For consumers, it enables a seamless checkout experience without ever leaving the AI interface. 

Despite the recent — and somewhat surprising — announcement that OpenAI is scaling back support for Instant Checkout in ChatGPT, ACP continues to gain traction across the broader ecosystem. As commercetools Founder and Chief Innovation Officer Dirk Hoerig puts it: “Standards like ACP are important because they create the connective tissue between AI agents and the systems merchants rely on to run their businesses.”

At the same time, OpenAI appears to be taking a more deliberate approach to stepping back, refining technical specifications, shifting toward webMCP as a framework for standardized AI–web interactions, and sharpening its focus on enterprise use cases.

For brands exploring how to stay visible, relevant and transaction-ready in the agentic era, implementing ACP specifications remains a practical step toward becoming agent-compatible in the near term.

Today’s commerce systems are built for humans clicking through websites — not AI agents. For merchants willing to transact on GenAI channels, standardizing how AI agents discover products, initiate checkout and complete purchases is crucial for success. 

ACP provides this “common language” (AKA, a standard that governs all communications between merchants and AI agents), acting as a secure bridge between agent and merchant systems. This enables businesses to effectively make their products available and shoppable on GenAI interfaces. 

Put simply, ACP outlines how AI agents can securely perform commercial tasks — such as making purchases, booking services, generating quotes and managing subscriptions — on behalf of users. It shifts AI from merely “finding” information to actively “doing” things.

These are the most relevant players in the ACP ecosystem:

• AI platforms: Providers like OpenAI that power the agents, interpret user intent and orchestrate actions across services.

• Payment providers: Companies such as Stripe and PayPal handling secure transactions, authentication and financial compliance.

• Commerce platforms: Platforms like Shopify and commercetools that manage product catalogs, pricing, inventory and order fulfillment.

• Merchants and brands: Businesses that expose their products and services via ACP-compatible interfaces.

By connecting AI platforms, payment providers, commerce systems and brands, ACP lays the foundation for a new era of agent-driven shopping, where products are not just discovered — they are actively purchased, booked and managed on behalf of users.

Consumer behavior began shifting from traditional eCommerce to AI-powered journeys, effectively opening new discovery channels for brands to reach high-intent customers. As zero-click commerce accounts for over a quarter of search engine volume, directly impacting how shoppers discover and consider products to purchase, brands and merchants need to become visible and transaction-ready in these channels. 

ACP is the layer that enables AI-driven shopping, making it possible to engage consumers in conversational journeys, such as autonomous quoting and replenishment orders: 

• “Book four round-trip flights from New York to Tokyo for next month and pick the best seats.”

• “Get me quotes for 1,000 branded T-shirts from our usual suppliers with next-day delivery.”

ACP handles product discovery, comparison and checkout orchestration for such scenarios safely and transparently, while enabling businesses to maintain their customer relationships as the merchant of record. 

That being said, this is what brands and retailers can do with ACP today:  

• Participate in conversational commerce: AI assistants are becoming increasingly conversational thanks to advances in AI models. Every conversation represents a potential commerce opportunity. ACP enables brands to seamlessly tap into these interactions and unlock new sales channels.

• Control what’s sold through AI assistants: ACP gives brands full control over which products are offered via AI assistants. Unlike traditional online marketplaces, businesses retain control over which products can be sold, how they’re presented and how orders are fulfilled. They can tailor product selections, feature specific brands or target particular product types to maximize conversions and align with your strategy.

• Maintain your role as merchant of record: ACP lets your brand leverage the reach and visibility of AI platforms without losing control. You remain the merchant of record, retaining ownership of customer relationships, pricing, inventory and retention strategies — not just a vendor on a marketplace.

• Enable agent-driven commerce: Make your product catalogs accessible to AI agents, support agent-initiated checkout flows and allow AI assistants to complete transactions on your behalf.

• Early examples of capabilities: Instant checkout within AI assistants, AI-driven product comparisons and recommendations, and fully conversational commerce experiences that guide users from discovery to purchase.

In addition, ACP is expected to support adaptable setups for all kinds of commerce, covering physical and digital products, subscription services and asynchronous transactions.

Early agentic implementations are promising, but the broader retail and commerce ecosystem is still evolving to fully support ACP. It’s important to understand where ACP stands today: 

Fully autonomous shopping is still early

Most purchase flows still require explicit user confirmation. AI agents can guide, recommend and initiate actions, but they aren’t yet completing orders entirely on their own. For instance, while 63% of European consumers report using AI tools for comparing options like brands, models, prices and reviews, only 21% report reordering previously purchased items or restocking household supplies. According to McKinsey, consumers are most comfortable with AI that assists decision-making while preserving final human control. 

Consumer trust in AI autonomy may increase over time as infrastructure matures, especially as brands design solutions around conditional delegation and reversibility, even for more autonomous use cases. 

Not a replacement for your storefronts

Agentic commerce is an emerging channel — complementary to your existing eCommerce platforms, not a substitute. Your current websites and apps remain central to your sales strategy. 

Not a payment provider

ACP doesn’t store or process sensitive payment information. Secure payments still rely on regulated providers, like Stripe and PayPal. AI agents pass payment tokens between parties without handling payment data directly.

Your business rules still live in your commerce engine

ACP is not a repository for your full business logic. Inventory management, post-purchase workflows and complex orchestration continue to be handled by your commerce backend — where platforms like commercetools excel. 

Brand control, visibility and loyalty questions remain

How brands influence AI-driven recommendations and maintain visibility is still evolving. Full transparency and control are not yet guaranteed. Finally, how AI-driven shopping on channels like ChatGPT will impact loyalty remains to be seen.

So, how does agentic commerce through ACP work in practice? Here’s a simple walkthrough of a standard ACP-enabled transaction process: 

Looking beyond ACP, there are other emerging protocols worth watching out for: 

• MCP (Model Context Protocol): A standardized framework for structured, secure context exchange between AI models and external systems. MCP defines how agents access tools and data (such as product catalogs, carts, or customer information) in a governed way, improving reliability and interoperability across systems.

• UCP (Universal Commerce Protocol): An emerging Google-led standard designed to enable interoperable commerce interactions between agents, merchants, and payment systems. UCP aims to standardize how product data, purchasing flows, and transaction signals are exchanged across platforms.

Agentic commerce requires an infrastructure that allows AI agents to securely access context, execute transactions and operate across multiple emerging standards. Rather than relying on a single protocol, enterprises should support a complementary stack of protocols that together enable discovery, checkout and system integration.

commercetools helps brands and retailers bridge their existing commerce infrastructure with agent-driven channels with the AI Hub. 

This solution provides a universal orchestration layer that translates complex brand logic into any AI channel (ChatGPT, Microsoft Copilot and Gemini) and, by extension, all emerging protocols, including ACP, MCP and UCP. 

The AI Hub enables businesses to: 

• Expose product catalogs to AI agents: Make product information — including pricing, availability and variants — machine-readable so agents can discover and recommend the right products.

• Manage checkout APIs: Support agent-initiated checkout flows, enabling flexible payment and shipping options while maintaining secure, standardized processes.

• Integrate with payment providers: Facilitate secure token-based transactions without storing sensitive payment data, leveraging trusted, compliant providers.

• Orchestrate order management: Connect commerce services from order creation through fulfillment, ensuring that inventory, confirmation,and post-purchase flows remain synchronized.

In short, commercetools provides agent-ready commerce APIs, flexible checkout flows, and service orchestration that together allow AI agents to act confidently on behalf of users, while the brand retains full control over pricing, inventory, and customer relationships.

Agentic commerce is evolving rapidly, and brands that prepare now will be best positioned to capture its benefits. Key actions include:

• Monitor the evolution of agentic commerce: Stay informed about emerging standards, early use cases and platform capabilities so you can adapt strategies quickly.

• Make your commerce infrastructure API-first: Agent-driven commerce relies on structured, machine-readable APIs for product discovery, checkout, and order orchestration. Ensure your systems expose capabilities programmatically.

• Ensure product data is structured and accessible: High-quality, well-structured product catalogs allow AI agents to accurately discover, recommend and transact products on behalf of users.

By focusing on these areas, brands can seamlessly integrate with AI assistants, maintain control over their commerce operations, and unlock new growth channels.