Product Hub MCP (Model Context Protocol) connects your Canto PIM data to an AI assistant, such as Claude, ChatGPT, or Microsoft Copilot. Once connected, you can search your product catalog, generate and update content, run bulk operations, and manage your data using plain-language prompts — without exporting files or editing records one at a time.

This article covers everything you need to get started and make the most of the connection:

• Part 1: Setup — prerequisites, generating your API key, and connecting your AI assistant
• Part 2: What you can do — use cases and example prompts once you're connected

Part 1: Setup

Before you start

To set up Product Hub MCP, you will need:

• An active Canto account with the Product Hub add-on enabled
• Product Hub Admin permissions in Canto (required to generate the API key in Step 1 — individual users do not need Admin permissions for Step 2)
• A supported AI assistant: Claude, ChatGPT, or Microsoft Copilot

How access works

Setting up Product Hub MCP involves two steps, performed by two different people:

• Your Admin generates a tenant-level API key (Step 1). This is done once for your entire organization.
• Each user connects their own AI assistant using that key, then authorizes with their own Canto login (Step 2). What each user can read or update through the MCP connection is governed by their existing Canto permissions — no new access controls are needed.

Step 1: Generate your API key

This step is performed once by a Product Hub Admin. Individual users do not need to repeat it.

1. Log in to Canto as an Admin.
2. Click your profile icon to open the Profile & Settings menu, then select Settings.
3. Open the Configuration Options tab.
4. Locate the API Keys section and click Create API Key.

Share the API key with users who need to connect their AI assistant in Step 2. You will also need your MCP server URL, which follows this format:

https://{your-subdomain}.canto.com/api/v1/pim/mcp

Replace {your-subdomain} with the part of your Canto URL that appears before .canto.com when you are logged in.

Step 2: Connect your AI assistant

Select your AI assistant and follow the steps for that platform.

Claude

Claude has native MCP support and provides the most complete Product Hub experience.

1. Go to claude.ai/settings/integrations while logged in to Claude.
2. Select Add custom integration or Add MCP server, depending on your Claude plan.
3. Fill in the following fields:
   • Name: Canto Product Hub (or any label you prefer)
   • MCP Server URL: https://{your-subdomain}.canto.com/api/v1/pim/mcp
   • API Key / Token: Paste the key generated in Step 1
4. Click Connect, then follow the prompt to authorize with your Canto credentials.

Once connected, the integration will appear in your active connectors list.

ChatGPT

1. In ChatGPT settings, navigate to Connectors or Custom integrations.
2. Add a new connection using the following details:
   • Endpoint URL: https://{your-subdomain}.canto.com/api/v1/pim/mcp
   • Authentication: API key (use the key from Step 1)
3. Authorize with your Canto credentials when prompted.
4. Save the connection.

Microsoft Copilot

1. Open Copilot Studio and navigate to Connections or Custom connectors.
2. Create a new connector with the following details:
   • Server URL: https://{your-subdomain}.canto.com/api/v1/pim/mcp
   • Authentication: OAuth / API key (use the key from Step 1)
3. Publish the connector to make it available in your Copilot environment.
4. Authorize with your Canto credentials when prompted.

Verify your connection

Once connected, start a new conversation in your AI assistant and try one of the following prompts:

• "List my Product Hub catalogs."
• "List all product categories."
• "Show me all product types and their attributes."

If your catalog data is returned, the connection is working correctly.

Permissions reference

What you can do through the MCP connection depends on your Canto user role:

Troubleshooting

The connection fails or times out

Check that the MCP server URL is entered exactly as shown, with no trailing slash. Check that the API key was copied in full with no extra spaces.

“Unauthorized” error after entering credentials

Your Canto account may not have API access enabled. Ask your Canto Admin to check your user settings under Admin Settings > User Management.

The connection succeeds but returns no data

Your Canto user role may have restricted read permissions. Contact your Admin to confirm your Product Hub access level.

The API key was lost or needs to be replaced

A Product Hub Admin can return to Admin Settings > Custom Configuration to generate a new key. All users with the old key will need to update their connector with the new one.

Part 2: What you can do

Once connected, you can manage your product catalog using plain-language prompts. This section covers the main use cases and includes example prompts to get you started. The prompts below are starting points — adapt the language to match your own catalog structure, attribute names, and workflows.

Generate and populate product content

Product content teams spend significant time writing descriptions, marketing copy, and localized text — drawing on fragmented inputs like SKUs, material lists, categories, and colors. The work is repetitive, output is inconsistent across team members, and products routinely go live with empty fields or placeholder text.

With Product Hub MCP, you can describe what you want in plain language. Your AI assistant reads your existing attribute data, generates content that fits your catalog structure, and writes results back to the correct fields — across hundreds or thousands of products in a single session. Teams shift from writing to reviewing.

Example prompts

• "Write a marketing description for every product in my Fall catalog that doesn't have one, using the name, material, and category."
• "Translate every English description in the Outerwear category into German."
• "Write a short description for all active products in the Footwear category using the name, color, and material attributes."

Find and fix content gaps

Incomplete product records are hard to spot until they cause a problem — a channel export fails, a storefront shows a blank field, or a buyer presentation reveals products with no copy. Without a reliable way to audit the full catalog, gaps only surface after something goes wrong downstream.

With Product Hub MCP, you can ask your AI assistant to audit your catalog against any set of attributes, auto-fill what can be generated, and return a clear picture of what still needs a human. You get actionable visibility at the moment it matters — before anything is published.

Example prompts

• "Show me every product in my Spring channel that's missing a description or a localized field, and fill in what you can."
• "List all products in the Outerwear category where the material attribute is empty."
• "Find all active products with no linked assets and no description."

Bulk create and update products from a file

Keeping Product Hub in sync with supplier files, planning sheets, agency hand-offs, and legacy exports means constantly handling data that wasn't prepared with your PIM in mind. Column names don't match your attributes, some rows are new products that need to be created, others are existing records that need specific fields updated, and required values are often missing or inconsistently formatted. Reconciling all of this manually is slow and error-prone at scale.

With Product Hub MCP, you can upload a file and describe what you want done. Your AI assistant reads the data, asks clarifying questions before making any changes, and then executes the right operation on each row — creating what's new, updating what exists, and flagging what it couldn't confidently resolve.

Example prompts

• "Sync this file with Product Hub."
• "Upload this supplier sheet — create any new products and update the existing ones."
• "I've uploaded a planning file with new Outerwear products. Create the new SKUs and update any existing ones you find."

Tips for writing prompts

When writing a prompt, try to include:

• The scope of products you want to work with — a catalog, category, product type, or specific SKU
• The attributes you want to read or update
• Any conditions that should apply — for example, "only products with no description" or "only active products"