Manual: Connecting Claude Desktop App to Icecat MCP

Version 0.9 (Updated on September 5th, 2025)

This guide provides step-by-step instructions on how to install and configure the Claude Desktop App to connect with Icecat MCP using an API Access Token.

Icecat MCP enhances LLMs with accurate knowledge about product data, thus avoiding the weak spots of AI: hallucinations. With Icecat MCP, you can be sure that a chatbot will provide accurate data.

1. Prerequisites

Before you begin, make sure you have a registered Icecat account with the channel partner user type.

2. Install Required Software

1. Download and Install the Claude Desktop App
2. Download and Install Node.js

3. Obtain Your Icecat API Access Token

1. Log in to your Icecat account.
2. Go to My profile settings.
3. Scroll down to the Access Tokens section.
4. Click on Manage Access Tokens.
5. Either:
   • Create a new API Access Token, or
   • Copy an existing one.

⚠️ Only use the Free Open Icecat account, given its permissive open content license. Don’t use a Full Icecat account, as we can’t be certain at this point in time that the AI respects proprietary data.

⚠️ Important: Keep your API token secure. Do not share it with unauthorized persons.

🔍 Access Tokens as an Alternative to IP Whitelists and app_key

4. Configure Claude Desktop App

You need to update a configuration file named claude_desktop_config.json.

1. Follow Settings > Developer > Local MCP Servers and click on Edit Config, and you will be redirected to the folder you need to place your config file in.

2. Update or create the configuration file in the Claude Desktop directory. It usually already exists as an empty JSON file.
3. ⚠️ Make sure the file extension is .json and not .txt.

Insert the following content into the file claude_desktop_config.json:

 {
  "mcpServers": {
    "icecat-mcp": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.icecat.biz",
        "--header",
        "Api-Token:YOUR_API_TOKEN",
        "--header",
        "Lang-Code:en"
      ]
    }
  }
}

1. Replace YOUR_API_TOKEN with the Icecat API Access Token.
2. Replace en in the Lang-Code line with the desired language code to get information for that locale. This parameter is optional. If not specified, 🇬🇧 EN will be used as the fallback locale. 🔍 Check the Icecat Locales and Language Code Table
3. Save the file.

5. Finalize Setup

1. Restart the Claude Desktop App. ⚠️ Make sure to close it also on the tray.
2. On the next launch, the app will automatically connect to Icecat MCP using your configuration.
3. You can verify that you are connected to the MCP by clicking on Settings:

Claude requires permission to use the external MCP and each of the tables; allowing it once means that each new message (not chat) pop-up will appear again.

---

6. Scopes

The main purpose of Icecat MCP is to enhance your Language models with accurate information about the products.

In the current stage, the AI agent must provide the following product information:

• ProductCode
• Brand

The MCP allows an AI agent to access the following information about the products:

• Product identifiers – The AI agent uses this tool when the user asks for basic product identifiers. Includes basic information: brand, name, product code, GTINs, and category.
• Product texts – The AI agent uses this tool when the user asks for descriptive or promotional product information. Includes: marketing text, official product naming, summaries, disclaimers, and bullet point highlights.
• Product specs – The AI agent uses this tool when the user asks for detailed product specs, features, and feature groups.
• Product overview – The AI agent uses this tool when the user asks for a general overview of available product details in one response. Includes basic information from all other tools.

In future versions, additional product data is to be included in the MCP.

Use-case

For example, you want to check what kind of Bluetooth is supported by the product Philips SHB9850NC/00. Instead of checking this manually, you can ask Claude.

With Icecat MCP, AI must check the data in our tables before sending it. Therefore, the chatbot will always provide accurate information about the product or clarify that Icecat does not have the product data.

An example of how Claude checks product specs before providing an answer to the customer’s question.

Check it out yourself

1. Ask to provide a full overview of a product.
2. Send two Product Codes and ask to compare their specs.
3. Ask how a certain product is positioned by supplier.

More about AI in Icecat

🔍 Icecat MCP server: Enhance your AI with World-Class Product Intelligence

🔍 AgenticFlow: how to integrate your own AI Agent with Icecat MCP

🔍 Mindpal: How to integrate your own AI Agent with the icecat MCP

🔍 How to integrate your BoltAI Desktop app with Icecat MCP