ACP implementation

A shopper describes what they need to an AI assistant, compares a few options, and decides to buy. Whether the purchase can be completed there, and whether it produces a valid order in your systems, depends on what your store can answer when the assistant asks.

The same applies in B2B. A purchasing team can instruct an agent to source parts that meet a given specification, compare terms and availability, and place the order within its approval rules. That only works if your catalog, contract pricing, and stock data are accurate and reachable.

What you keep is as important as what you gain. With delegated payment handling, payment credentials pass through your existing payment service provider. Your fraud rules, 3D Secure checks, decline handling, and retry behavior stay in place, and you remain the merchant of record, so the customer relationship and the transaction stay yours.

The Agentic Commerce Protocol is an open standard, co-developed by OpenAI and Stripe and published under an open source license, that defines how an AI agent and a merchant exchange cart, payment, and order information so a purchase can be completed on a shopper's behalf. It can be implemented as a REST interface or as an MCP server. The standard itself is free. The cost sits in connecting it to the systems that already run your store.

ACP defines the exchange. It does not define your commerce logic. Tax configuration, returns and exchanges, fraud modeling, and how your payment provider authorizes and captures a charge all stay where they are today, which means your systems have to return the correct answer at the moment an agent asks for one.

That is also why a ready-made connector rarely finishes the job. Variants may come from a PIM. Available stock may sit in an inventory or ERP system. The final price may depend on discounts, fulfillment choices, or tax rules evaluated during checkout. We map ACP operations to the systems that already control those decisions rather than replacing them, on Magento, WooCommerce, Umbraco, Shopify, and custom builds.

Some platforms offer ready-made plugins that let AI agents communicate with a store. But these plugins are built for a standard setup. They don't automatically understand how your business handles product data, volume pricing, stock, or a customized checkout. Connecting those parts is where most of the work lies.

We implement UCP as a layer over the systems already running your store rather than replacing them, drawing on years of experience in building and extending Magento, WooCommerce, Umbraco, Shopify, and custom ecommerce systems. That covers both the implementation work on platforms without native support, and the data and configuration work Shopify stores need instead.

UCP can connect your store to AI agents in more than one way. It can use standard web connections such as REST or MCP, which is a common way for AI systems to connect to tools and data. We do both sides of that connection. Read about our MCP server development, which makes your store available to approved agents, and our AI agent development, which builds agents that use it.

• Checkout session endpoints: Create, retrieve, update, complete, and cancel, mapped to your store's actual cart and order behavior.
• Product feed export: Catalog data structured for ACP and mapped from your real product attributes, variants, and inventory fields.
• Delegated payment handling: Payment routed through your existing provider, with its fraud checks, 3D Secure, declines, and retries intact.
• Capability declaration and extensions: A statement of what your store supports, including discounts and fulfillment where they apply.
• Order webhooks: Signature verification, idempotency keys, retry handling, and monitoring for duplicate or delayed events.
• Checkout-time resolution: Current tax, stock, and pricing supplied by your own systems at the moment an agent requests checkout.
• Monitoring and alerting: Visibility into checkout session failures, webhook failures, and product feed errors.

1. Start with a fixed-price technical audit. We begin with a technical audit of your ecommerce stack. The resulting capability matrix shows which ACP requirements are available now, available with configuration, require development, or are blocked by a system limitation, and which payment route applies to you. That last point matters early, because whether your provider already supports an agentic route or your store needs a custom handler affects cost and timeline more than any other choice in the project.
2. Move from findings to implementation. Project Scoping turns the audit findings into an agreed scope and price for a Fixed Price Project. If your team wants to shape the solution hands-on with our developers, Co-Dev is the alternative.
3. Keep the specification current. After launch, Continuous Monthly Development can cover specification upkeep, including assessing new releases and updating the implementation when needed.