What Is the Agentic Commerce Protocol? The Open Standard Powering AI Shopping

Last Updated: October 2025

What Is the Agentic Commerce Protocol (ACP)?

The Agentic Commerce Protocol (ACP) is an open-source standard that enables AI agents to facilitate purchases on behalf of users. Developed by Stripe and OpenAI, ACP creates a common language between businesses, AI platforms, and payment processors—allowing seamless transactions directly within AI conversations.

Key announcement: On September 29, 2025, OpenAI launched Instant Checkout in ChatGPT, powered by ACP. The system enables 700 million weekly ChatGPT users to purchase products directly in conversation—from discovery to checkout to fulfillment—without ever leaving the chat interface.

The protocol is:

• Open-source under Apache 2.0 license
• Platform-agnostic - works with any AI agent or payment processor
• Merchant-friendly - businesses maintain customer relationships and control
• Flexible - supports physical goods, digital products, subscriptions, and asynchronous purchases

ACP solves a fundamental coordination problem: before this standard existed, every merchant would need custom integrations with every AI platform. ACP standardizes the handshake between agents, businesses, and payment systems—build once, connect everywhere.

How Does Agentic Commerce Protocol Work?

ACP orchestrates three-way interactions between buyers, AI agents, and businesses:

The Purchase Flow

1. Discovery and Recommendation
User asks their AI assistant for product recommendations. The AI queries available merchants, evaluates options based on user preferences, and presents relevant choices.

2. Product Selection
User selects a product. The AI retrieves detailed information from the merchant's ACP-enabled checkout endpoint, including pricing, availability, shipping options, and terms.

3. Secure Payment Authorization
Rather than exposing payment credentials directly, ACP uses tokenized payment credentials scoped to specific merchants and transaction amounts. The AI agent securely passes payment tokens between buyer and merchant.

4. Transaction Completion
The merchant processes the payment through their existing payment infrastructure. The AI agent confirms the transaction and provides order details to the user.

5. Order Fulfillment
The merchant handles fulfillment, shipping, and customer service through their existing systems. The business retains the customer relationship as merchant of record.

Technical Implementation

Merchants implement ACP by:

• Publishing checkout configuration via API or Model Context Protocol (MCP)
• Creating agent-facing endpoints that integrate with existing commerce backends
• Configuring payment token acceptance through compatible payment service providers
• Defining which products can be sold through agentic channels
• Setting business rules for transaction acceptance (per-agent, per-transaction, or custom logic)

For Stripe merchants, enabling agentic payments requires as little as one line of code using Stripe's Shared Payment Token API.

Why Was the Agentic Commerce Protocol Created?

The E-Commerce Evolution

Commerce has shifted through distinct waves:

• Physical stores to e-commerce websites
• Desktop shopping to mobile commerce
• Manual browsing to personalized recommendations
• Click-to-buy to voice commerce

AI represents the next fundamental shift: agent-mediated transactions where AI systems handle product research, comparison, and purchasing on behalf of users.

The Integration Problem

Before ACP, enabling AI-driven commerce required:

• Custom integrations between each merchant and each AI platform
• Separate payment authorization flows for every agent
• Platform-specific product data formats
• Unique fraud prevention implementations per integration
• Different security models across platforms

This fragmentation made it operationally impossible to scale agentic commerce. Every new merchant or AI platform required bespoke development.

The Solution: Open Standards

Stripe and OpenAI spent a year building ACP to create standardization that:

• Works across platforms - any AI agent can transact with any ACP-enabled merchant
• Uses existing infrastructure - merchants don't rebuild their commerce systems
• Maintains business control - merchants decide what's sold and how it's presented
• Enables secure payments - tokenized credentials prevent payment data exposure
• Supports any payment processor - not limited to Stripe (though Stripe has first-mover advantage)

What Are the Benefits of Implementing ACP?

For Merchants and Businesses

Reach 700 Million Weekly ChatGPT Users
OpenAI's Instant Checkout provides immediate access to hundreds of millions of high-intent buyers. Early implementation means appearing in AI-driven shopping results before competitors.

Reduce Cart Abandonment
Traditional e-commerce sees 60-80% cart abandonment rates. Agentic commerce eliminates friction—users go from product discovery to completed purchase in a single conversation, often in under a minute.

Maintain Customer Relationships
Unlike marketplace models where platforms own the customer relationship, ACP ensures businesses remain the merchant of record. You control product presentation, fulfillment, customer service, and post-purchase engagement.

Use Existing Infrastructure
ACP integrates with your current commerce backend, payment processor, and inventory systems. No need to rebuild or migrate to new platforms.

Pay Only for Completed Transactions
Merchants pay a small fee on completed purchases. The service is free for users, doesn't affect pricing, and doesn't influence product recommendation rankings.

For AI Platforms

Embed Commerce Without Becoming a Merchant
AI platforms can enable transactions without handling payments, inventory, fulfillment, or customer service. Businesses retain these responsibilities while AI platforms facilitate discovery and checkout.

New Revenue Streams
Transaction fees from completed purchases create monetization beyond subscriptions or advertising, aligning platform incentives with successful commerce.

Enhanced User Experience
Users complete tasks entirely within the AI interface—no context switching, no opening browsers, no re-entering payment information across different merchants.

For Payment Processors

Process Agentic Transactions at Scale
As commerce moves from websites to AI workflows, payment processors that support ACP become the infrastructure layer for this new channel.

Secure Token-Based Payments
ACP's token architecture creates new security models for delegated purchasing, positioning compatible processors for the next generation of commerce.

Who Is Using the Agentic Commerce Protocol?

Current Implementations

OpenAI / ChatGPT
Instant Checkout launched September 29, 2025, with Etsy sellers. Over a million Shopify merchants—including Glossier, SKIMS, Spanx, and Vuori—are being added. ChatGPT acts as the user's AI agent, securely passing information between user and merchant.

Stripe
First payment processor to support ACP with Shared Payment Token API. Stripe handles payment authorization, fraud detection, and transaction processing while maintaining merchant relationships.

Etsy
U.S.-based Etsy sellers can now receive purchases directly through ChatGPT conversations, accessing hundreds of millions of potential buyers without additional marketing spend.

Shopify Merchants
Over a million Shopify stores are being integrated, making their products available for purchase through ChatGPT's Instant Checkout feature.

Expected Adoption

Google released its own Agent Payments Protocol (AP2) backed by 60+ merchants and financial institutions, signaling competitive validation of agentic commerce infrastructure.

Microsoft opened its Copilot Merchant Program, indicating plans for commerce integration across its AI platforms.

Anthropic / Claude has not announced ACP adoption, but the protocol's open-source nature makes integration straightforward if Anthropic chooses to enable commerce features.

How Do Businesses Implement the Agentic Commerce Protocol?

Implementation Options

Option 1: Stripe Integration (Fastest)
If you're already on Stripe:

1. Update to Stripe's latest SDK
2. Enable Shared Payment Token API (often one line of code)
3. Configure your product catalog for agent access
4. Apply for ChatGPT Instant Checkout inclusion

Option 2: Custom ACP Implementation
For businesses using other payment processors:

1. Review ACP specification at agenticcommerce.dev
2. Implement checkout configuration endpoints
3. Integrate with compatible payment service provider's ACP support
4. Publish product data in agent-readable format
5. Configure business rules for transaction acceptance

Option 3: Platform Integration
Many e-commerce platforms will offer native ACP support:

• Shopify is integrating ACP for its merchants
• Other platforms will likely follow as adoption grows
• Check with your e-commerce platform for their ACP roadmap

Technical Requirements

Product Data Structure
AI agents need rich, structured product information to make recommendations:

• Detailed specifications and attributes
• High-quality images and descriptions
• Accurate pricing and availability data
• Shipping options and timelines
• Return policies and warranty information

The more complete and structured your product data, the better AI agents can match products to user needs.

Real-Time Inventory
ACP works best with live inventory data. Out-of-stock recommendations damage trust and user experience.

Programmatic Pricing
Enable dynamic pricing rules that agents can query programmatically, including discounts, promotions, and volume pricing.

Fulfillment Integration
Automated order processing and fulfillment tracking improve the agent-mediated shopping experience.

What Are the Challenges of Agentic Commerce?

Loss of Brand Presentation Control

Traditional e-commerce gives businesses complete control over how products are presented—visual merchandising, storytelling, emotional appeal. In agent-mediated commerce, the AI agent decides how to present your product to users.

Businesses must shift from visual and emotional branding to structured, verifiable differentiation that agents can process programmatically. Product attributes, specifications, certifications, and reviews become more important than creative marketing.

Discovery Algorithm Concerns

How do AI agents decide which products to recommend? OpenAI states that Instant Checkout items are not preferred in results and that rankings consider availability, price, quality, whether a merchant is the primary seller, and whether Instant Checkout is enabled.

But as agentic commerce grows, questions about recommendation transparency, potential bias, and merchant visibility become more critical. The algorithm driving recommendations becomes as important as Google's search algorithm.

Reduced Direct Customer Interaction

When agents handle transactions, businesses have fewer touchpoints for building customer relationships. The pre-purchase discovery and decision-making process happens inside the AI conversation, not on your website.

Businesses must find new ways to differentiate, build loyalty, and create value beyond the transactional moment.

Payment Security Concerns

Users may hesitate to provide payment information to AI agents, even with tokenized credentials. Building trust in agent-mediated transactions requires transparent security practices and clear communication about how payment data is protected.

Multi-Item Cart Complexity

ACP currently supports single-item purchases through ChatGPT. Multi-item carts, complex configurations, and bundled offerings require additional protocol development.

How Does ACP Compare to Traditional E-Commerce?

Traditional E-Commerce Flow

1. User searches Google or browses directly to merchant website
2. User navigates product pages, reads descriptions, views images
3. User adds products to cart, reviews items
4. User proceeds to checkout, enters shipping and payment information
5. User completes purchase

Average time: 15-45 minutes
Touchpoints: Multiple websites, search engines, review sites
Cart abandonment: 60-80%

Agentic Commerce Flow

1. User asks AI agent for product recommendation
2. Agent queries available merchants, presents relevant options
3. User selects product
4. Agent handles checkout with stored payment credentials
5. Purchase complete

Average time: 1-5 minutes
Touchpoints: Single conversation
Cart abandonment: Significantly lower (estimated 20-40%)

The efficiency gains are dramatic. The merchant relationship dynamics fundamentally change.

What Happens Next for Agentic Commerce Protocol?

Near-Term Evolution (2025-2026)

Multi-Item Carts
OpenAI has indicated that multi-item cart support is coming to ChatGPT's Instant Checkout. This enables more complex purchase patterns and higher average order values.

Geographic Expansion
Initial launch is U.S.-only for both merchants and buyers. International expansion will unlock significantly larger markets.

Additional Merchant Categories
Beyond Etsy and Shopify, expect restaurant delivery, service bookings, digital product sales, and subscription offerings to adopt ACP.

Competitive Protocol Adoption
Google's AP2 and Microsoft's Copilot Merchant Program will drive innovation and potentially create interoperability challenges as multiple standards compete.

Long-Term Transformation (2026+)

Agent-to-Agent Commerce
AI agents negotiating with other AI agents on behalf of their users—comparing prices, requesting customization, coordinating bulk purchases, and handling complex procurement.

Personalized Pricing and Offers
Merchants offering AI agents dynamic pricing based on user history, loyalty status, purchase patterns, and inventory optimization needs.

Automated Replenishment
AI agents proactively reordering consumable products based on usage patterns, inventory depletion, and predictive needs.

Service Integration
Expanding beyond product purchases to service bookings, reservations, appointments, and complex multi-step transactions.

Should Your Business Adopt the Agentic Commerce Protocol?

ACP makes sense when:

You sell products online and want access to hundreds of millions of potential buyers using AI-powered shopping assistants.

You can provide structured product data with detailed specifications, accurate pricing, and real-time inventory information.

You're willing to compete on product merit rather than purely on marketing and brand presentation, since AI agents evaluate products programmatically.

You have efficient fulfillment operations that can handle increased order volumes without compromising delivery timelines or customer service.

You use or can integrate with compatible payment processors that support ACP's tokenized payment approach.

ACP might not make sense when:

Your products require extensive consultation or customization that doesn't translate well to agent-mediated transactions.

Your competitive advantage is experiential rather than product-based (though physical showrooms can complement agent-mediated online sales).

You lack technical resources to implement structured product data and maintain integration quality.

Your products are new-to-market and benefit from educational content and brand storytelling that agents can't effectively communicate.

Key Takeaways: Agentic Commerce Protocol in 2025

The Agentic Commerce Protocol represents a fundamental shift in how online transactions occur. By standardizing the interaction between AI agents, businesses, and payment systems, ACP enables commerce to happen wherever users are—inside AI conversations, without leaving the interface.

OpenAI's September 29 launch of Instant Checkout in ChatGPT moved agentic commerce from theoretical to operational, with 700 million weekly users now able to make purchases through AI-mediated transactions.

The protocol is open-source, platform-agnostic, and designed to work with existing commerce infrastructure—making adoption more achievable than complete platform migrations.

However, agentic commerce changes the rules: businesses must compete on structured data quality and product merit rather than visual merchandising and emotional branding. Direct customer relationships evolve as agents mediate the discovery and decision-making process.

Early adopters gain access to massive new customer bases while learning how to optimize for agent-driven discovery. Late adopters risk invisibility as more shopping moves into AI interfaces where traditional web presence doesn't matter.

The commerce infrastructure shifted from websites to agents faster than most businesses expected. The question isn't whether to adopt—it's how quickly you can execute.