Understanding the Norce Backend MCP Servers

This article explains what the Norce Backend MCP Servers are and how Norce Commerce is structured — in plain language, without technical detail. If you are a PIM manager, pricing manager, or operator who wants to get the most out of an AI assistant connected to Norce, this is the right place to start.

For connection and setup instructions, see Using the Norce Backend MCP Servers.

What the MCP servers are

When you connect an AI assistant — such as Claude — to Norce Commerce, you are not giving the AI a window into the Admin UI. Instead, the AI gets access to a set of purpose-built tools. Each tool knows exactly how to read or update a specific part of Norce: one tool lists price lists, another checks product readiness, another imports stock quantities.

This is what the MCP servers are: a collection of those tools, organised by domain. There is one server for product data, one for pricing, one for inventory, one for supplier management, and one for general configuration. When you describe what you want, the AI decides which tools to use, calls them in the right order, and presents the result in plain language.

You do not need to know which tool does what. That is the AI's job. Your job is to describe what you are trying to find out or accomplish.

It is safe to explore

Read operations are always safe — the AI can query anything without changing anything. Write operations pass through a built-in guard: lower-risk changes go through with a visible warning, and higher-risk changes on stage and production are refused until explicitly confirmed. See the Security and access guide for a full explanation of how write protection works and what it does not cover.

You can chain questions

The AI remembers what it found during your conversation. If you ask it to check a product and it finds a pricing issue, you can immediately follow up with "why is that happening?" or "show me the same product on the Norwegian storefront" without starting over. This is what makes working through an AI different from running individual searches in the Admin UI — the context carries forward.

Why this helps you

Once you have this map in mind, your prompts become more specific — and more specific prompts get better answers.

Instead of: "Why is the price wrong?" Try: "The Norwegian price for product AB-1042 looks too low. Check the raw stored price and the Norwegian storefront preview side by side and tell me where the difference comes from."

Instead of: "Why isn't this product showing up?" Try: "Product AB-2201 is not appearing in the Lighting category on the Swedish storefront. Does it have a primary category assigned, the Published flag set, and is it active?"

The AI can interpret vague questions, but specific ones get faster and more useful results — and knowing the shape of the data is what makes specificity possible.

How Norce Commerce is structured

To ask good questions, it helps to understand what Norce keeps track of and how the pieces connect. You do not need to know all the technical details — but having a rough map makes your prompts more specific and your results more useful.

Products and SKUs

In Norce, a product is the parent — the thing you sell. A SKU (also called a variant) is a specific, purchasable version of that product. A t-shirt is a product; the t-shirt in size M, colour blue, is a SKU. Most tools work at the SKU level, identified by a PartNo — the unique code for each specific variant.

Parametrics

Parametrics are the attributes that describe a product — colour, weight, material, voltage, dimensions. They can be different types: a list of fixed values (colour: Red, Blue, Green), a number (weight: 1.5 kg), a boolean (yes/no), or free text.

Parametrics are what make products findable and filterable in the storefront. A product missing a required parametric value might not appear in the right filters, or might be blocked from publication.

Flags

Flags are simple on/off markers assigned to products. Common examples: Published, Clearance, Featured, Marked for outlet sales, No discounts allowed. Flags are used to control visibility, apply campaign pricing, trigger business rules, and filter products for specific purposes.

A product with no flags assigned at all is often incomplete — flags are a common thing to check before publication.

Price lists

Prices in Norce are not stored directly on products. They live in price lists — named collections of prices that can be assigned to different storefronts, sales areas, and customer groups. A product can have different prices on different price lists.

Price lists can also have rules — for example a campaign price list that automatically applies a 20% discount on top of the standard retail price list.

Warehouses and stock

Physical stock is stored in warehouses, and within warehouses in locations. When a customer checks availability, Norce applies business rules on top of the raw stock numbers — which warehouses to include, what thresholds to apply, how to calculate the displayed quantity.

This means there are two views of stock: the raw number stored in the database, and the availability number shown to a customer. They can differ, and understanding which one you are looking at matters when investigating a stock issue.

Suppliers

Suppliers are the organisations that provide products to the client. In Norce, each supplier has its own warehouses, locations, and cost price lists — separate from the client's retail warehouses and price lists. Supplier products are mapped to client SKUs by manufacturer part number.

Applications

An application in Norce represents a specific channel — a storefront, a brand, or a sales context — with its own configuration: which price lists it uses, which sales area it belongs to, which availability rules apply. A single Norce environment can have multiple applications for different reasons. Sometimes they represent different markets: a Swedish storefront and a Norwegian storefront. Sometimes they represent different brands or business models selling to the same market — for example a consumer brand like "All Spice SE" and a wholesale brand like "Spice Imports Inc", each with their own pricing, assortment, and customer base.

Almost every tool call in the Backend MCP Servers requires an application ID. This is because the answer to many questions — what is the price of this product?, how much stock is available? — depends on which storefront you are asking from.

The same product can have different prices, different availability, and even different content depending on which application ID you use. This is not a bug — it is how multi-market commerce is designed. Understanding it helps you ask more precise questions: not just "what is the price?" but "what is the price on the Swedish storefront?"

Common questions

Do read operations change any data?

No. All read operations — looking things up, checking values, comparing data — are safe to run at any time. The AI queries data but does not create, update, or delete anything unless you explicitly ask it to and confirm the action.

Does the AI have access to my Norce Admin UI?

No. The AI connects to Norce Commerce through the Backend MCP Servers, which use Norce APIs. It cannot see or interact with the Admin UI in any way. Everything it can do is through the structured tools the MCP servers expose.

What data does the AI send to my AI provider?

The content of your conversation — including prompts and tool results — is sent to and handled by your AI provider (such as Anthropic, Microsoft, or GitHub). Tool results from the MCP servers, which contain Norce Commerce data, are part of that conversation. If your organisation has specific data residency or privacy requirements, review your AI provider's data policies before using it with production data. See also the Security and access guide.