4.9

MCP app development for eCommerce brands

Your catalog and order data available inside ChatGPT and Claude, so customers can check stock and reorder without leaving the assistant.

Trusted by 700+ leading brands worldwide

Trusted by 700+ leading brands worldwide

What an MCP app is

An MCP app is a service that puts your business inside an AI assistant. It is built on the Model Context Protocol, an open standard that lets ChatGPT and Claude read live data and carry out actions in external systems, which means a shopper can see pricing and reorder without leaving the chat.

Read the full technical guide in What is MCP.

Commerce systems we already connect

800+

eCommerce integrations built

6+

Commerce platforms supported

10–15

Days to a live Launchpad

Why merchants are building MCP apps now

Shoppers are already asking assistants what to buy. Without an MCP app, an assistant can only repeat what it found on your website, and it cannot see a live price or a customer's account.

Stale answers about your catalog

Assistants describe your products from pages they crawled months ago, including items you no longer stock and prices you have since changed.

No account, no repeat order

A returning customer cannot see their contract pricing or rebuild a previous cart, so the assistant sends them back to the website to start again.

Competitors answer first

When an assistant can complete the task inside the chat for one brand and not another, the brand it can act for is the one recommended.

Book an MCP app call

What your customers can do inside the assistant

An MCP app shows a defined set of actions, and you decide which ones. Each is connected to the system that owns the data, so an answer in chat matches what your store would show.

Live stock and pricing

Availability and price read from your commerce platform at the moment of the question, including promotions and customer-specific rates.

Account recognition

A signed-in customer is matched to their account, which gives the assistant their order history and any contract pricing you apply.

Reorder and cart building

Customers rebuild a previous order or add items to a cart in chat, then finish in your checkout with everything already applied.

Quotes and B2B approvals

Buyers request a quote against their contract terms, and the request enters the same approval flow your sales team already works in.

Returns and order status

Delivery status and return requests are answered from your order system, without the customer opening a ticket or waiting for an agent.

Support from your customer service

Product and policy questions are answered using your existing helpdesk content, so the assistant repeats what your support team would say.

Which of these workflows should your first ChatGPT App run?

Send us your platform and the questions your customers ask most. We'll come back with the workflow to build first, the systems it connects to, and a realistic timeline.

Three productized ways to get a live MCP app

Most brands begin with Launchpad to prove the connection, then extend once the internal case is made.

Launchpad
MCP Launchpad server
Live in 10–15 business days
A production MCP server connected to your catalog and stock, with a defined set of read actions. It proves the connection works against your real data and gives your team something to test inside ChatGPT or Claude.
Production-ready MCP server
Catalog and stock actions
Tested in ChatGPT and Claude
Source code and setup docs you own
Start with Launchpad
Pilot
ChatGPT app pilot build
A submitted ChatGPT app, live in 4–6 weeks
A customer-facing app inside ChatGPT, with account recognition and the actions your buyers use most. Includes the interface work and the submission process, plus analytics on what your customers actually ask for.
Everything in Launchpad
Account and order actions
Submission and analytics
Custom widgets and account sign-in
Book a Pilot walkthrough
Suite
Commerce AI App Suite
Multi-market and B2B scope, live in 8–12 weeks
The commerce surface across ChatGPT and Claude, including B2B account flows and multi-market rules. Built for brands treating the assistant as a real sales channel, with the connector packs for your ERP or PIM.
Everything in the Pilot
B2B and multi-market flows
ERP and PIM connectors
Live in ChatGPT and Claude
Book a Suite walkthrough

How an MCP app build is delivered

Every package follows the same sequence, with the difference in how far it goes. The server is built first, because nothing else can be tested until the assistant can read your real data.

Scope the actions

We agree which questions and tasks the app should handle, and which systems hold the data behind each one.

Build the MCP server

The server is built on our MCP Core and the connector pack for your platform, then connected to your catalog and order systems.

Design the in-chat experience

The cards and flows customers see inside the assistant are designed against your brand using the Widget Kit, so replies are visual.

Submit and launch

We handle the submission process, then monitor the first weeks of live traffic and tune the actions according to what people ask.

Built by commerce engineers

We already own the integrations

Exposing live commerce data safely is the hard part of an MCP build, and it is the same work our teams do on Adobe Commerce and Shopify every week.

Pre-built connector packs

MCP Core, the commerce connector pack, and a Magento 2 and Hyvä module already exist, which is what keeps the first build to a fixed price.

Certified for enterprise data

Customer accounts and order history are in scope under the same ISO 27001, ISO 27017, and PCI DSS certifications as the rest of our delivery.

Assistant-agnostic by design

One MCP server serves both ChatGPT and Claude, which makes adding another assistant later a configuration step on work you already own.

The server stays yours

You own the MCP server and its connectors outright, so no vendor platform ends up holding your commerce data or your customer accounts.

Designed for the chat

Replies appear as branded cards built with our Widget Kit, giving customers something to tap and act on inside the assistant itself.

What clients say about working with scandiweb

scandiweb is a supportive, solution-oriented partner. We like working with you guys. Proactive and reactive. Great quality.
Claudia Schioenning
Head of eCommerce
Partnering with scandiweb’s team allowed us to bring a site to market that has 3 – 4 times better performance than any other site.
Jason Barney
eCommerce Technology Consultant
For more than 10 years, scandiweb 
has supported our platform with top talent, helping us reach our strategic goals.
Jonathan Chan
Head of Global IT
We have been thrilled to work with your team and hope to continue long-term—you have been doing fantastic work!
Steve Mayne
Executive Commerce Director
scandiweb is our strategic partner 
for end-to-end development and 360° eCommerce expertise, including UX and data.
Henri Kruusel
Head of eCommerce & Marketing
Working with scandiweb on our platform has been a pleasure. They are always trying to find the best solutions.
Marc Muntané
eCommerce Manager
scandiweb has consistently demonstrated exceptional expertise, adaptability, and a proactive approach.
David Bethune
Senior eCommerce Manager
It’s been an extremely fruitful relationship and we are really, really happy.
Jeanine Frutuoso
Director of Marketing
I value scandiweb as a partner and very happy to recommend them! I would very much like to keep the existing team for the long term.
Steve Keatley
CTO at Beauty Works

Curious what a ChatGPT App would look like for your business?

Tell us your platform and what your customers ask for most. We'll come back with an app idea built on your stack, the systems it connects to, and a realistic timeline. No commitment.

MCP app questions and answers

Do we need an MCP app if we already do SEO?

They solve different problems. AI search optimization gets your brand mentioned when someone asks an assistant for a recommendation. An MCP app lets that person then check their pricing and buy without leaving the chat. Most brands need the visibility first and the app once that traffic is worth converting.

Which commerce platforms do you support?

Adobe Commerce, Shopify, BigCommerce, commercetools, Salesforce Commerce, SAP Commerce, and WooCommerce, plus the ERP, CRM, PIM, and OMS behind them. A Magento 2 and Hyvä module already exists. For other systems we build the connector as part of the project.

Who maintains it after launch?

We do, under a monthly support package. Assistants change their interfaces and rules often, and your catalog changes constantly, so an app left alone degrades. Support covers monitoring, action tuning against real questions, and updates when the assistant platforms change.

Which assistants does an MCP app work in?

ChatGPT and Claude both support MCP, and one server serves both. ChatGPT reaches the larger consumer audience through its Apps SDK, while Claude is more common in B2B and internal use. We build the server once and configure it per assistant, so the second one costs a fraction of the first.

How much does an MCP app cost?

MCP Launchpad is €9,900 and delivers a working server in 10–15 business days. A ChatGPT App Pilot is €29,900 over 4–6 weeks. The Commerce AI App Suite starts at €74,900 over 8–12 weeks. Ongoing support starts at €4,900 a month once the app is live.

How long before customers can find it?

A server your team can test is the fastest milestone. A customer-facing app also has to pass the assistant platform's submission review, which adds time outside our control. We build the submission into the Pilot and Suite packages and manage it for you, including the changes reviewers ask for.

Talk to the team building commerce MCP apps

Tell us which platform your store is on and what you would want a customer to be able to do inside ChatGPT or Claude. We will come back with the package that fits and the actions it would cover.

Call with a solution architect
Free action and data-access map
Fixed-price package

Prefer to talk now? Book a call straight away, or email us at: [email protected]

We reply within one working day. Your details stay private.

More on MCP and AI apps

Related AI services

Find out what your MCP app would let customers do

We will review your platform and data access, then tell you which package fits and what it would cover.