Norce Backend MCP Servers bring AI closer to the operational side of commerce. They give commerce teams a new way to work with their actual Norce environment – using AI to investigate, validate and act across complex commerce data and workflows. Now available in open beta, they are an important step towards making AI a practical part of everyday commerce operations.
What if you could ask your commerce backend a question in plain language and not only get an answer based on what is happening in your Norce environment, but also act on it? With Norce Backend MCP Servers, your AI assistant can use structured tools to investigate, validate and, with the right permissions and safeguards, make changes.
A product has the wrong price in your Norwegian storefront.
Where do you start?
Is the source price wrong? Is a price list configured incorrectly? Is VAT being applied differently? Is it a rounding rule? Or has the latest change simply not propagated to the storefront yet?
Finding the answer can mean moving between different parts of Norce Admin, checking several APIs or asking a developer to investigate.
Now, you can simply ask:
“Show me the raw stored value and the storefront preview side by side, and tell me where the difference comes from.”
And it doesn't have to stop at the answer. Depending on the task and the permissions you've granted, your AI assistant can use the Backend MCP tools to take action too.
This is the idea behind the new Norce Backend MCP servers, now available in open beta.
The Backend MCP Servers provide a structured connection between AI applications and the Norce Commerce APIs.
Instead of an AI assistant simply telling you how Norce works, it can use defined tools to retrieve information from your actual Norce environment and, with the appropriate permissions and safeguards, perform actions.
The servers use Model Context Protocol (MCP), an open standard that enables AI applications to connect to external systems and tools.
This means you can connect Norce Commerce to supported AI clients and work conversationally with your commerce backend.
You don't need to memorise tool names or look up every application ID manually. Describe what you want to investigate, and the AI can determine which tools it needs to use, call several when necessary and correlate the results.
The important distinction is that this isn't a chatbot answering questions from documentation.
When you ask for a product's price, availability or configuration, the underlying values are retrieved through the same Norce APIs used by other integrations.
Your AI assistant becomes a new way of working with the commerce engine you already have.
The current open beta includes five domain-specific Backend MCP Servers:
Config covers applications, stores, sales areas, VAT codes, currencies and store configuration.
Product covers products, categories, variants, parametrics, bundles, relations, translations and product content.
Pricing covers price lists, pricing rules, calculated prices, rounding, imports and best-price history.
Inventory covers warehouses, locations, stored stock and customer-view availability.
Supplier covers suppliers, supplier warehouses, cost price lists and supplier product feeds.
Together, they expose around eighty structured tools as of August 2026.
The servers can work independently, but the real potential becomes apparent when a question crosses several domains.
Ask why a product isn't available at the expected price, for example, and the AI can investigate product status, pricing and inventory as part of the same conversation.
One particularly useful capability is the ability to compare two different views of your commerce data.
The raw view tells you what is actually stored in the backend, before VAT, business rules, rounding or caching.
The customer view shows what the storefront receives after those rules have been applied.
Putting those two views side by side can make troubleshooting significantly easier.
Instead of only discovering that something is wrong, you can investigate where the difference originates.
For an ecommerce team, that can turn questions such as “Why is the site showing this?” from a cross-team investigation into a conversation with the platform.
The opportunity isn't limited to troubleshooting.
Commerce teams often need to answer questions that traditional administration interfaces aren't designed to answer efficiently across a large assortment.
Imagine preparing to launch a German storefront. Instead of checking categories and product metadata individually, you could ask:
“Check what's missing for German (de-DE) before we launch the German storefront. Group the results by entity type.”
Or before launching a campaign:
“Show me what prices these products will have when the campaign becomes active.”
A pricing manager can inspect future-dated campaign pricing before customers see it. A PIM manager can identify missing translations or incomplete products across an assortment. An ecommerce manager can compare the same product across different storefronts. And a developer can inspect configuration across stage and production without writing another one-off lookup script.
The interface is conversational. The underlying work is still structured.
Some tasks shouldn't have to be explained from scratch every time.
That's where Skills come in.
A Skill is a saved playbook that tells the AI how to perform a specific workflow consistently. The current setup includes workflows for investigating prices, checking pricing health, checking product readiness, finding incomplete products and importing data from files.
For example, the investigate-price Skill can compare the lowest stored price across active public price lists with the storefront result and help identify whether the difference comes from data, business logic, VAT, rounding, sales area or propagation.
But these Skills are just a starting point.
Customers and partners can create and customize their own Skills around the workflows, business rules and challenges that matter to them. Combine that flexibility with the tools available through the Backend MCP Servers, and the possibilities become much broader than the examples we've created so far.
This is where we believe things get really interesting.
We're providing the tools and some initial playbooks, but we expect our customers and partners to discover use cases we haven't even thought of yet. New ways to investigate issues. New ways to validate data before launch. New ways to streamline recurring tasks and bring AI into existing commerce processes.
We believe this can be a game changer for how people work with commerce, not because AI replaces the expertise of the people running it, but because it gives that expertise a new way to interact with the platform.
For teams working with large and complex assortments, this moves AI beyond ad hoc questions and answers towards repeatable, business-specific commerce workflows.
And we're genuinely excited to see what our customers and partners build next.
There is an equally important boundary to understand.
Backend MCP Servers aren't designed to replace direct API integrations.
Scheduled integrations, continuous data synchronisation, high-volume automation and real-time production monitoring still belong in purpose-built integrations using the Norce APIs.
MCP solves a different problem.
It gives people – and the AI tools they choose – a structured, conversational way to investigate, understand and work with Norce Commerce.
That distinction is important. Composable commerce has always been about using the right capability for the right job. AI should be no different.
As Norce expands its agentic capabilities, you'll encounter several different MCP offerings.
They serve different purposes.
Norce Backend MCP acts. It works with the real configuration, products, prices, inventory and supplier data in your Norce environment.
Norce Assistant MCP explains. It understands Norce documentation, APIs and the data model, but doesn't read or change your commerce environment.
Norce Commerce MCP powers customer-facing commerce experiences, such as product discovery, baskets and conversational shopping.
A useful way to think about the first two is simple:
The Backend servers act. The Assistant server explains.
Connect them alongside each other and your AI assistant can understand both the Norce model and your actual environment.
We believe AI will change more than the storefront.
It will change how commerce teams interact with the infrastructure behind it.
For years, we've built Norce Commerce around openness, APIs and composability. MCP is a natural extension of that strategy: instead of locking AI into another proprietary interface, we make the capabilities of the commerce engine available to the tools our customers and partners choose to work with.
The Norce Backend MCP Servers are currently in open beta, and we're actively developing them based on feedback from customers and partners. New tools and capabilities will continue to be added, with additional domains planned for areas including baskets, orders, promotions and post-purchase management.
This is still the beginning.
But you don't need to start by handing AI the keys to your commerce operation.
Start with a question.
Connect your stage environment. Run a check. Compare what is stored with what your customer sees.
And see what happens when your commerce backend becomes part of the conversation.
Connect to Norce via Claude desktop