Advice that helps.
An assistant that knows the catalogue and belongs to the merchant’s own shopping experience.
Shopware · Agentic Commerce Lab
We build so merchants can shape what comes next.
Agents are changing how people find products, compare offers and buy. At Shopware’s Agentic Commerce Lab, we turn that shift into concrete projects, developed in the open with merchants’ interests at the centre.
A team at Shopware. A public space for experiments, references and new possibilities.
Why the Lab exists
The next purchasing journey
is taking shape.
We work where the next product generation is still taking shape. How does an agent discover a merchant’s offer? How does a brand remain tangible? When may an assistant act? Those questions become prototypes, protocol work and tools that others can try and develop further.
Our reference point is commerce itself: more helpful purchasing, greater freedom to build a business and decisions that remain with the merchant. What proves useful in an experiment can become the foundation for an integration or a product.
How we learn in the Lab · Chapter 17The brand comes along.
Product cards and relevant experiences make an offer tangible in a conversation with an assistant.
The shop can respond.
Explicit tools make products, variants and the cart available to agents.
Intent becomes an order.
Discoverable capabilities, reliable offers and clear handoffs connect the purchasing journey.
The projects
Explore them.
Try them.
Build on them.
The public repositories show what we are working on. This page connects the ideas; each project contains its code, documentation and current development status.
01
Discover & connect
An agent first needs to understand what a merchant offers and which capabilities the shop provides.
Lab project
Agentic Resource Discovery
Help an agent understand what a shop offers and how to work with it. The plugin publishes a directory of data, services and available actions.
A starting point for a purchasing journey in which capabilities can be discovered before an agent uses them.
Lab project
Agentic Experience Protocol
A product is more than a name and a price. AXP brings variants, configurations, quality signals and embedded experiences into an agent’s interface.
It explores how advice and the distinctive qualities of an offer can matter in an agent channel.
Research preview
WebMCP for Shopware
Search products, select variants and work with the cart: the shop exposes these capabilities inside the current browser session.
Explicit tools connect the assistant to the shopping journey the customer can see on the website.
02
Advice & experience
Helpful advice and a relevant interface connect product knowledge to customer intent. In the shop and in the assistant.
Research preview
Shopping Assistant Starter Kit
A shopping assistant in the merchant’s own shop: customers can ask questions, compare products and add suitable items to their cart.
Merchants and developers adapt the model, product knowledge, capabilities and presentation to their business. Purchasing continues through the regular Shopware checkout.
Lab project
Sales Agent Harness
A sales agent needs reliable rules. This framework connects advice, product search and cart preparation to the merchant’s policies and systems.
Prices and commercial commitments come from the connected commerce system. The same flow can serve chat interfaces and agent-to-agent communication.
Lab project
Shopware as an MCP App
Bring the catalogue into the assistant: customers search for items, see product cards and prepare their cart in a conversation.
An embedded interface connects the conversation to the checkout where the customer completes the purchase.
Lab project
A2UI for Shopware Frontends
What does a shop look like when its interface responds to the customer’s current intention? This experiment lets an agent generate relevant views and render them in the browser.
It connects A2UI, Shopware Frontends and a product interface assembled live in the browser.
Lab project
Page Agent Bridge
A customer describes what they want to do. An agent panel can read the shop page and carry out steps through its interface.
It explores another way into the existing storefront experience: natural language and visible interaction.
03
Offers & operations
An answer becomes an action: an offer, a prepared purchase or a change in the business. The merchant sets the boundaries.
Alpha
Merchant Quote Agent
A B2B customer asks for a better price. The Quote Agent responds within the discount limits set by the merchant.
The history shows offers and decisions. Requests beyond the pricing remit go to the team; draft mode lets people approve every response first.
Lab project
Shopware × Claude Commerce
Two perspectives on the same business: a shopping agent finds products and prepares a purchase. A merchant agent reviews sales and proposes changes.
Price, stock and promotion changes are checked in a preview and then approved by the merchant.
Research preview
x402 Payments
How does an automated buyer pay? This project explores a payment flow in which the shop sends a payment request and the buyer signs it digitally.
The reference connects Shopware orders to the x402 protocol and documents local tests and payments with test tokens.
04
Build & verify
Ideas become software more quickly. Our tools help put agents to work deliberately and understand their results.
Lab project
Pawl
Agents can write a great deal of code. Pawl connects requirements to verifiable outcomes and records which work actually passes a check.
A tool for development and research that makes progress visible through verified effects.
Lab project
Agent in a Box
A workspace for the coding agent: tools and runtime live in a development environment dedicated to the project.
It helps prepare environments deliberately and run agents separately from the everyday workstation.
Lab project
RFQ Bench
Which agent negotiates better, and at what cost? RFQ Bench compares buyers, sellers and strategies in controlled B2B negotiations.
Recorded traces make offers, walk-away limits and outcomes inspectable and available for fresh analysis.
Lab project
Lightcone
A game of hide-and-seek with an agent that sees very little of its surroundings. It needs to use tools and adapt its search during play.
A playful experiment in perception, action and the information an agent actually needs.
Inside the projects
Where an idea becomes
a visible workflow.
Two original captures from the public project documentation, showing how offers, decisions and approvals connect.
Offers with an inspectable history.
The Merchant Quote Agent shows the counteroffers and the path from a request to an order. This capture comes from the documented demo shop.
A proposal. A preview. An approval.
In the Claude Commerce experiment, the merchant sees the price change prepared by the agent. The portal connects analysis to a visible approval.
Research & talks
Ideas to
build on.
Not every question starts with a plugin. Research reports and talks connect practical work to the wider shift.
Research report
After Turing
How can we distinguish people, bots and agents? This research report examines behaviour, environment and metadata as possible signals.
The report, sources and figures are published alongside the typesetting project.
Talk
Shopware Commerce Day 2026
Stefan Hamann and Daniel Nögel on the practice of AI-assisted software development: the keynote as a complete presentation readable in the browser.
With the examples and visual explanations from “A Reality Check on AI-Driven Software Engineering”.
From experiment to product
The UCP reference has taken its next step.
The early Lab project led to an integration into Shopware’s official Agentic Commerce plugin. The original repository remains available as a reference.
Project overview checked on 15 October 2026, based on the Shopware Agentic Commerce Lab’s public repositories and documentation. About the Lab