Agentic Commerce with Shopware: How to get your online shop ready for AI shopping!

Your customers are no longer searching for products just via Google, marketplaces, or your online shop's navigation. More and more frequently, they are simply describing what they need to an AI assistant:

“Find me a waterproof outdoor jacket in size M for a maximum of 180 euros that can be delivered by Friday.”

The AI assistant can research matching offers, compare products, and take into account price, features, availability, and delivery conditions. Depending on the platform and technical integration, it can then even put together a shopping cart, prepare a checkout, or accompany an order.

Exactly this shift is known as Agentic Commerce. For merchants, this creates a fantastic new contact and sales channel – but only if product data and shop processes are also understandable and accessible for machines.

In this context, two technical building blocks are often mixed up: Product feeds and the Universal Commerce Protocol (UCP) are not the same thing. A feed transmits product data to a platform. UCP allows compatible AI agents direct, controlled interaction with the shop.

In this post, we'll show you how Agentic Commerce works, what Shopware already provides, and how you should prepare your shop today!

Agentic Commerce in a nutshell: Product feeds ensure that AI platforms know your range and can find products. Protocols like UCP also create a standardized connection to the shop, through which compatible agents – depending on your approvals – can access the catalog, shopping cart, checkout, or orders.

What is Agentic Commerce?

Agentic Commerce refers to the use of AI agents in e-commerce. Unlike a classic chatbot, these systems don't just answer individual questions. They pursue a goal and can combine several steps to achieve it.

For example, a chatbot explains which jacket is suitable for rainy weather. A shopping agent can go beyond that to research suitable products, compare offers, and prepare the purchase process.

The decisive difference lies in taking action: The agent doesn't just process information, but can – provided the technical connection and permissions allow it – interact with real commercial processes.

For Shopware merchants, this creates two exciting tasks: Your range must be discoverable and understandable for AI platforms. And if an agent is to do more than just research, it needs controlled access to shop functions.

Product Feed and UCP: Two Different Tasks

At this exact point, a clear distinction is highly worthwhile. In its documentation on the Agentic Commerce extension, Shopware explicitly lists the Product Feed Sales Channel and UCP as two separate components.

The Product Feed: So AI Platforms Know Your Products

At its core, a product feed is a structured product list. Shopware provides its own Agentic Commerce sales channel for this purpose and currently supports, among others, an OpenAI product feed as well as a Google product feed.

Details such as product name, description, price, availability, images, and variant information are transmitted. The external platform reads or imports this data and can use it for product searches and recommendations.

A concrete example: You export 20,000 items from Shopware. An AI platform processes the feed and subsequently knows which products you offer, what they cost, and whether they are available.

The feed makes your product range visible – but it does not automatically give the agent the ability to create a shopping cart in your shop or execute an order.

OpenAI also describes product feeds as the basis for indexing products in ChatGPT and displaying them in shopping experiences. The data must be updated regularly for this. OpenAI: Product Feeds

UCP: Enabling an Agent to Interact with Your Shop

The Universal Commerce Protocol (UCP), on the other hand, is an open standard for communication between AI agents, platforms, and commerce systems. It doesn't just describe product data, but functions along the entire shopping journey.

Through UCP, Shopware can, for example, enable the following capabilities:

• Query product catalog,

• Create and modify shopping carts,

• Apply discounts,

• Perform checkout processes,

• Create and retrieve orders,

• Link identities and process tokenized payment information.

For this, the shop publishes a machine-readable UCP profile. In it, a compatible agent can see which capabilities and interfaces are actually available. You decide what an agent is allowed to do – not the agent itself!

Simply put: The feed says "These are my products." UCP says "This is what you can do in my shop with these products."

Does UCP Replace the Product Feed?

No, at least not across the board. Although UCP can also provide catalog information, meaning there are technical overlaps, feeds and UCP play different roles in current platform integrations.

For example, Google continues to describe existing Merchant Center product feeds as the foundation for product discovery, while UCP enables agentic actions and direct purchases on AI interfaces. Google: Universal Commerce Protocol

Shopware also treats both paths separately: The Agentic Commerce Product Feed distributes product data to external platforms; UCP provides approved shop functions for compatible agents.

For merchants, this means: Product feed and UCP are not competing solutions, but rather two complementary building blocks of the same architecture.

And What About ChatGPT? UCP is Not ACP

Here, too, the terminology is slightly confusing. OpenAI documents the Agentic Commerce Protocol (ACP) for ChatGPT. The product feed provides catalog data for search and shopping; for the checkout embedded in ChatGPT, OpenAI describes its own ACP processes. OpenAI: Agentic Commerce – Key Concepts

So, UCP and ACP are two different protocols. Shopware's extension currently supports product feeds for OpenAI and Google, and alongside that, UCP as an open agent standard. If you want to implement "Agentic Commerce with ChatGPT," you should therefore not automatically assume that an activated UCP interface already establishes the ChatGPT checkout.

Your Online Shop Becomes a Machine-Readable Sales Area

No matter which channel a product is found through: The quality of the underlying data remains absolutely crucial.

An AI agent requires reliable information such as:

• Product name and an easy-to-understand description,

• Price, currency, and availability,

• Variants like color, size, or material,

• Brand, identifiers, and product features,

• Images and accessible product URLs,

• Delivery, return, and cancellation conditions.

If this information is incomplete or contradictory, an agent will have a harder time classifying your offer – even if the product actually fits the query perfectly.

Product data is thus becoming a second, machine-readable storefront. Targeted optimization of your online shop therefore not only helps your human visitors, but also improves the conditions for AI-based product recommendations!

What Does Shopware Offer for Agentic Commerce?

Shopware now bundles several components with which merchants can analyze their shop, distribute product data, and provide agent functions in a controlled manner.

Agentic Readiness Scanner

With the Agentic Readiness Scanner from Shopware, you can get a great first impression of how well your store is prepared for AI shopping. The scanner looks at accessibility, product data, trust signals, and new agent standards, among other things.

It serves as a fantastic starting point, but does not replace a detailed check of your actual product data, processes, and system landscape.

Agentic Commerce (Beta)

The free Shopware extension Agentic Commerce (Beta) currently bundles both technical paths:

• Product Feed Sales Channel: generates product feeds for external platforms like OpenAI and Google.

• UCP: publishes selected sales channels and capabilities for compatible agents.

With the feed, you control parameters such as the range of products, variants, language, currency, updates, and tracking. With UCP, on the other hand, you define which functions of the shop an agent is allowed to use. Depending on the version, Shopware supports various transmission paths such as REST, A2A, Embedded, and MCP.

This distinction is particularly important: Activating a feed does not automatically open the checkout for agents. And activating UCP does not automatically replace the feed-specific connection to an AI platform.

Shopware Intelligence and Copilot

In Shopware, Agentic Commerce does not just affect external shopping agents. Shopware Intelligence and Copilot also bring AI features directly into the administration. Merchants can use them to analyze shop data, answer questions, or get support with multi-step tasks.

This is a different use case than the Agentic Commerce sales channel: Here, the agent works for the merchant, while external shopping agents research and act for the customer.

How Does a Purchase via an AI Agent Work?

An agent-based purchase can be broken down into five simple steps:

1. Formulate the wish: The customer specifies product features, budget, or delivery date.

2. Find products: The agent uses available product data, feeds, and other shared sources.

3. Evaluate offers: Price, stock, variants, delivery time, and conditions are compared.

4. Shopping cart and checkout: With the appropriate integration, the agent can interact with the shop and prepare the purchase.

5. Confirm and accompany order: The customer gives the necessary approval; the shop and platform then process the order and status information according to the respective integration.

The agent's scope of action is limited by technical permissions, business rules, and approvals.

What Is Particularly Important for Shopware Merchants Today?

For most merchants, the biggest short-term lever is not to immediately enable a fully autonomous checkout. What's much more important is a rock-solid foundation!

1. Set Up Product Data Cleanly

Titles, descriptions, variants, properties, prices, and stock levels should be complete, up-to-date, and unique. If ERP, PIM, or inventory management are the leading systems, the data must arrive reliably in Shopware.

2. Build Product Feeds Strategically

Decide which assortments should be visible on which AI platforms. Check the respective field requirements and ensure that price and availability are updated regularly.

3. Don't Just "Turn UCP All the Way Up"

If you are testing UCP, it should be clear beforehand which capabilities an agent really needs. Catalog access is very different from shopping cart, checkout, or order access. This is precisely why Shopware offers a configurable capability structure as well as security mechanisms for communication.

4. Think About Tracking From the Very Start

New AI channels can only be evaluated economically if hits, referred orders, and revenue can be accurately attributed. In the Agentic Commerce channel, Shopware supports affiliate and campaign codes as well as statistics on referred orders, customers, and sales.

How Far Along Is Agentic Commerce Already?

Status: August 6, 2026. The Shopware extension is still explicitly in the beta phase. The range of functions and behavior can change with future updates.

External platforms are also not available to the same extent everywhere. For example, Shopware currently points out that registration for the ChatGPT Marketplace is only possible in the USA. OpenAI also describes the onboarding of Product Feeds and Instant Checkout as currently available only for approved partners.

For merchants in Europe, this doesn't mean they have to wait and see! Product data, feeds, tracking, and system architecture can already be prepared today. UCP can be tested in a controlled manner – without having to approve every transactional access right away.

Our Assessment: Discovery First, Transaction Second

For most Shopware merchants, we would currently approach Agentic Commerce in two stages.

Stage 1 is Visibility: Is your product data structured so well that AI platforms can understand, compare, and recommend your assortment? Do feeds work reliably, and can new visitors and sales be measured?

Stage 2 is Interaction: Which agents should actually be allowed to talk to your shop? Which functions should be accessible via UCP or other platform-specific protocols? How will checkout, permissions, identity, and security be mapped?

This order prevents technical complexity from arising before the most important foundation is in place: reliable, high-quality product data.

How Nimbits Supports You with Agentic Commerce

For us, Agentic Commerce doesn't start with installing an extension. What matters is how product data, Shopware, ERP/PIM, business processes, and external AI channels play together.

We support you in making your shop agent-ready step-by-step – from analyzing product data to OpenAI and Google feeds, right through to UCP configuration, security, and success measurement. In doing so, we also check which functions are truly useful today and which are better activated once platforms and processes are fully ready for them.

With ongoing Shopware support, we ensure that your data, extensions, and interfaces continue to run smoothly long after the initial setup.

Frequently Asked Questions About Agentic Commerce with Shopware

What is the difference between a product feed and UCP?

A product feed transmits structured product data to an external platform. Its primary purpose is to make products discoverable and comparable. UCP, on the other hand, is a standardized interface for direct interaction between compatible agents and the shop, and can map shopping carts, checkout, and orders in addition to catalog queries.

Does UCP replace the OpenAI or Google product feed?

No. Current integrations treat product feeds and UCP as separate building blocks. Google continues to use existing Merchant Center feeds for product discovery; Shopware also offers separate OpenAI and Google feeds alongside UCP.

Are UCP and OpenAI's ACP the same thing?

No. UCP stands for Universal Commerce Protocol. OpenAI uses the Agentic Commerce Protocol (ACP) for its Agentic Commerce infrastructure. Both aim to technically enable agent-based commerce, but they are different protocols.

Is Agentic Commerce with Shopware available in Germany?

The Shopware extension can already be installed and tested. However, individual functions of external platforms may be restricted regionally or to approved partners. Therefore, you should always check separately what Shopware technically supports and what the respective AI platform has actually unlocked in your target market.

How do I make my Shopware shop agent-ready?

Start with clean product data and clear business rules. Next, set up the required product feeds and establish tracking. You should only activate and test UCP functions in a targeted manner – matching the processes that agents are actually permitted to use.