Understand what MCP means for your store
Get a clear, practical explanation of how Model Context Protocol connects AI agents to ecommerce systems and why it matters for your platform choice.
MF
AIMCP: How AI Agents Connect to Your Ecommerce Platform
Model Context Protocol is the emerging standard for connecting AI agents to external systems. For ecommerce, MCP determines whether AI assistants can browse your catalogue, check stock, and place orders. This guide explains what MCP means for your platform choice and technical architecture.
Fits with
What Is Model Context Protocol?
Model Context Protocol (MCP) is an open standard created by Anthropic and adopted by OpenAI, Google, and Microsoft. It defines how AI models connect to external tools, databases, and APIs. Think of it as a USB standard for AI: instead of building a custom integration for every AI assistant, you expose your system through MCP and any compatible agent can interact with it.
For ecommerce, this is significant. An AI shopping agent that supports MCP can query your product catalogue, check real-time pricing and stock, add items to a cart, and initiate checkout. It does this through structured API calls, not by scraping your website. The protocol handles authentication, data formatting, and error handling in a standardised way.
MCP was open-sourced by Anthropic in late 2024. Within a year, it reached 97 million monthly SDK downloads and over 10,000 active servers. The Linux Foundation now stewards the protocol, which removed enterprise procurement concerns and accelerated corporate adoption.
Why Ecommerce Merchants Should Pay Attention
The shift is already underway across every layer of the commerce stack. Shopify launched four official MCP servers covering storefront, customer accounts, checkout, and developer tools. Norce released a Commerce MCP Server with product search, cart management, and B2B-ready context handling. Shopware is developing its own MCP server and founded the Agentic Commerce Alliance with 15 partners to establish open standards. Salesforce Commerce Cloud and Commercetools have followed with their own MCP products.
The ecosystem extends beyond platforms. Adyen has an official MCP server for payment integration. Klaviyo launched an MCP server for campaigns, flows, and customer data. Storyblok offers a native MCP server with full content management capabilities. Visa, Mastercard, PayPal, and Stripe have all launched MCP-compatible payment tools for agentic commerce.
These are not experimental features. They are production infrastructure. When a customer asks an AI assistant to find and buy a product, the assistant uses MCP to interact with the store, the CMS, the payment provider, and the marketing platform. Each system that supports MCP becomes part of the agent's workflow. The infrastructure decisions you make now determine whether your store participates.
How MCP Works in Practice
MCP operates on a client-server model. The AI agent is the client. Your ecommerce system is the server. The server exposes a set of tools (functions the agent can call) and resources (data the agent can read).
A typical ecommerce MCP server might expose tools for searching products by attributes, checking inventory for a specific SKU, retrieving shipping options and estimated delivery times, adding items to a cart and initiating checkout, and looking up order status and tracking information.
The agent calls these tools based on the user's intent. If someone says "find me a blue wool sweater under 800 kr in size M with delivery before Friday," the agent translates that into structured API calls against your MCP server. Your server returns structured data. The agent evaluates the results and either completes the purchase or asks the user for a decision.
Authentication happens through OAuth 2.0 or API keys. The protocol supports granular permissions so you can control what external agents can access and what actions they can perform.
Platform Readiness for MCP
Norce
Norce has taken a strong position in MCP adoption. The Norce Commerce MCP Server is a production-ready server that exposes product search, product detail retrieval, cart creation, item management, and quantity updates through structured MCP tools. The server sits as a thin layer on top of Norce Commerce Services, reusing the same OAuth 2.0 authentication and supporting full B2B context: customer-specific pricing, company accounts, multiple price lists, and sales area targeting.
In addition to the Commerce MCP Server, Norce offers an Assistant MCP Server that gives AI tools access to Norce documentation and developer knowledge, and an Agent SDK for building custom commerce agents. This makes Norce one of the most MCP-ready platforms in the Nordic market.
Shopify
Shopify leads in breadth of MCP adoption. Its four official MCP servers (Storefront, Customer Account, Checkout, Dev) cover the full purchase journey. Shopify merchants get MCP compatibility without custom development. The Storefront MCP server exposes product search, pricing, and availability. The Checkout server handles cart management and payment initiation. Shopify also launched Agentic Storefronts, connecting 5.6 million merchants to AI assistants including ChatGPT and Perplexity.
Shopware
Shopware is actively building its MCP infrastructure. An official Admin MCP server (shopware-admin-mcp) is available on GitHub, giving AI assistants direct access to product management, category management, and theme configuration through the Admin API. Shopware is also developing a Commerce MCP server as a core technology for AI-to-shop data exchange.
Shopware founded the Agentic Commerce Alliance in 2025 together with 15 digital companies to establish open standards for agent-to-agent commerce and protect merchant independence. Shopware is also integrating with PayPal's Commerce Agent Platform, making merchant product data discoverable and purchasable through consumer AI experiences.
Magento / Hyvä
Magento with Hyvä has a growing MCP ecosystem. Community-built MCP servers are available on the Adobe Commerce Marketplace (Freento MCP Server) and GitHub (magenable/magento2-mcp). Adobe has committed to supporting UCP and ACP agentic commerce protocols, making merchant catalogues, pricing, and inventory machine-readable by AI agents. Adobe Experience Manager also supports MCP endpoints for content management.
The tradeoff for Adobe Commerce: full agentic AI capabilities require a separate Adobe Experience Platform (AEP) licence, which can significantly increase total software investment. For merchants on Magento Open Source or standard Adobe Commerce, the community MCP servers provide a practical starting point.
MCP Across the Commerce Stack
MCP adoption is not limited to ecommerce platforms. The tools and services surrounding your store are also becoming MCP-ready, creating an interconnected ecosystem where AI agents can orchestrate entire commerce workflows.
Adyen offers an official MCP server for payment integration, allowing AI agents to interact with payment processing through structured tools. Klaviyo launched its MCP server in 2025, giving AI agents access to campaigns, customer flows, profiles, events, and reporting data. Storyblok provides a native MCP server with full CRUD capabilities across stories, assets, components, releases, and workflows.
This means an AI agent building a commerce workflow can query your product catalogue (via Norce or Shopify MCP), check the customer's profile and segment (via Klaviyo MCP), personalise content (via Storyblok MCP), and process payment (via Adyen MCP). Each interaction follows the same protocol, reducing integration complexity and enabling multi-step agent workflows that span your entire stack.
What Your Store Needs to Be MCP-Ready
MCP readiness is not about installing a plugin. It is about ensuring your commerce infrastructure can serve structured, reliable data to non-human consumers.
Structured product data is the foundation. Every product needs consistent attributes, accurate stock levels, and machine-readable specifications. If your catalogue relies on freeform text descriptions without structured fields, AI agents cannot parse it reliably.
Real-time APIs are essential. Agents need current pricing and availability. Cached data that updates once daily creates a poor agent experience. Invest in real-time synchronisation between your ERP system and your commerce platform.
Programmable checkout means the ability to manage a cart, apply promotions, select shipping, and complete payment through API calls without a browser session. This is where many stores fall short.
Authentication and authorisation determine what external agents can do. OAuth 2.0 flows with granular permissions let you control access to customer-specific pricing, wholesale terms, and order management.
MCP and Agentic Commerce
MCP is the technical layer that enables agentic commerce. While agentic commerce describes the business shift (AI agents acting as buyers), MCP describes the technical standard that makes it work. Understanding both is important for planning.
The practical implication: investing in MCP readiness also improves your store for traditional channels. Better structured data improves search engine visibility. Better APIs improve integrations with ERP systems, CRM platforms, and marketplaces. Better checkout flows improve conversion for human buyers too. MCP readiness and operational maturity go hand in hand.
Getting Started
You do not need to build an MCP server from scratch. Many of the tools in a typical Nordic ecommerce stack already offer MCP support.
- Audit your current stack for MCP support. Check whether your platform (Norce, Shopify, Shopware, Magento), CMS (Storyblok), payment provider (Adyen), and marketing platform (Klaviyo) already offer MCP servers. Many do.
- Improve your product data. Ensure all products have structured attributes, consistent categorisation, and accurate inventory. This is the single highest-impact investment regardless of MCP.
- Connect the MCP servers you already have. If you run Norce with Storyblok, Klaviyo, and Adyen, you may already have four MCP-ready systems. The work is connecting them to an agent framework and testing the workflow end to end.
- Talk to Nordic Web Team. We work across all four platforms and the broader partner ecosystem. We can map which parts of your stack are already MCP-ready, identify gaps, and plan a pragmatic path to agent-enabled commerce.
Relevant systems in this setup
These systems often show up when we plan ecommerce for this type of business. Use them as concrete tracks for CRM, payments, and ERP.
Strengths
MCP protocol expertisePlatform-specific readiness assessmentAPI architecture guidanceAgentic commerce preparation
Business benefits
Assess your platform's MCP readiness
Learn where Shopify, Shopware, Norce, and Magento stand on MCP support and what each platform requires to serve AI agent traffic.
Identify technical gaps early
Map your current API surface, product data quality, and checkout programmability against what MCP-compatible agents require.
Invest in improvements that compound
MCP readiness improvements (better data, stronger APIs, programmable checkout) also benefit search visibility, integrations, and human conversion rates.
Delivery approach
MCP touches your entire commerce stack. Product data, pricing, inventory, checkout, and fulfilment all need to be accessible through structured APIs for AI agents to interact with your store. The protocol itself is a connector layer, but readiness depends on your platform architecture and data quality.
Beyond the integration
The integration is only one part of the work. Platform choice, data quality, content, UX, QA, and the launch itself also need to be planned and delivered for the solution to work in practice.
1
Audit your API surface
Map current API capabilities against what an AI agent needs: product queries, pricing, cart management, checkout, and order status.
2
Improve product data structure
Ensure all catalogue data uses consistent attributes, structured specifications, and real-time inventory synchronisation.
3
Evaluate MCP implementation path
Determine whether your platform offers native MCP support, whether community servers exist, or whether custom development is needed.
4
Build and test incrementally
Start with read-only product data exposure, then expand to cart management and checkout. Test with real AI agents to identify gaps.
FAQ
What is Model Context Protocol and who created it?⌄
MCP is an open standard for connecting AI models to external systems. Anthropic created and open-sourced it in 2024. OpenAI, Google, and Microsoft adopted it within months. The Linux Foundation now stewards the protocol.
Does my ecommerce platform already support MCP?⌄
Norce has a production Commerce MCP Server with product search, cart management, and B2B context support. Shopify has four official MCP servers covering the full purchase journey. Shopware has an official Admin MCP Server for product, category, order, and theme management. Magento has community-built MCP tools available on the Adobe Commerce Marketplace. All four platforms can serve AI agents, but the level of out-of-the-box support varies.
Is MCP the same as having a good API?⌄
MCP builds on top of APIs. A good API is a prerequisite, but MCP adds a standardised way for AI agents to discover, authenticate with, and interact with your system. It is the difference between having a door and having a door with a standard lock.
How much does it cost to become MCP-ready?⌄
For Norce and Shopify merchants, production MCP servers already exist. Shopware has an Admin MCP Server that can be deployed immediately. For Magento, community tools are available. The biggest cost across all platforms is usually improving underlying data quality and API coverage, which is valuable regardless of MCP.
Should I wait until MCP adoption grows before investing?⌄
The technical foundations for MCP readiness (structured data, comprehensive APIs, programmable checkout) improve your store today. Meanwhile, Stripe, Klarna, Visa, and Mastercard are all building agent payment infrastructure. The ecosystem is moving fast enough that waiting means catching up rather than evaluating.