MCP Setup Guide for Claude Desktop, Cursor, and Windsurf

The Model Context Protocol (MCP) lets you connect your YourGPT chatbot as a knowledgebase inside tools like Claude Desktop, Cursor, and Windsurf.

This guide help you through two main parts:

1. Setting up MCP inside YourGPT

2. Connecting it with your tool (Claude Desktop, Cursor, or Windsurf)

1. Configure MCP in YourGPT

YourGPT has two kinds of MCP:

Chatbot Knowledgebase MCP — this guide. Lets Claude, Cursor, or Windsurf search one chatbot's trained knowledgebase.

Platform MCP — lets an AI assistant manage your YourGPT account: create projects, add and remove training, read conversations, manage contacts, pull analytics. Jump to the section at the end of this guide.

Step 1: Open the MCP Integration

1. Go to YourGPT Dashboard:

2. Then Integrations → MCP (Model Context Protocol)

3. Click Configure.

Step 2: Create MCP Server

Click on Create MCP to generate your chatbot’s MCP configuration.

Step 3: Select Your Target Tool

After the server is created, you’ll see a configuration screen with tool-specific tabs for:

• Claude Desktop

• Cursor

• Windsurf

Each tab gives you a JSON block to copy, along with instructions on where to place it.

Part 2: Connect MCP with Your Tool

Now that your chatbot is MCP-ready, here’s how to connect it with each tool.

2.1 Claude Desktop

Step 1: Open Claude Settings

Click Claude in your menu bar and select Settings.

Step 2: Enable Developer Mode

From the settings sidebar, go to Developer and click Edit Config.

Step 3: Locate Configuration File

On macOS, go to:

~/Library/Application Support/Claude/claude_desktop_config.json

If it’s not there, relaunch Claude or check permissions.

Step 4: Add Your MCP Configuration

Open the config file in any text editor or IDE, and paste the code provided in YourGPT.

Paste the configuration block provided by YourGPT. It should look like this:

{
  "mcpServers": {
    "yourgpt_mcp": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.yourgpt.ai/sse",
        "--header",
        "Authorization: Bearer YOUR_KEY"
      ]
    }
  }
}

Step 5: Restart Claude and Confirm

Save your changes, Restart Claude Desktop.

If MCP is working correctly, you’ll see a hammer icon in the bottom-right corner.

2.2 Cursor Setup

Cursor setup is much simpler.

1. Look for Cursor Settings, then click on MCP, the add new global MCP server

2. Locate or create the file:

~/.cursor/mcp.json

3. Paste your MCP config JSON into this file.

4. Save and restart Cursor.

5. Test Your MCP.

2.3 Windsurf Setup

For Windsurf, follow the same approach.

Step 1: Locate the MCP Config File

~/.codeium/windsurf/mcp_config.json

Step 2: Paste Your Configuration

Use the config you got from YourGPT. It should look like this:

{
  "mcpServers": {
    "yourgpt_mcp": {
        "command":"npx",
        "args": [
            "mcp-remote",
            "https://mcp.yourgpt.ai/sse",
            "--header",
            "Authorization: Bearer YourKEY"
        ]
    }
  }
}

Paste it into mcp_config.json using your editor.

Step 3: Refresh Windsurf MCP & Test

Open Windsurf and look for the MCP section in the right panel. Click on Refresh.

Once it connects, your AI will appear under “available MCP servers”.

Manage your whole YourGPT account with Platform MCP

The setup above gives your AI assistant read access to one chatbot's knowledgebase.

If you want it to actually run your YourGPT account — create projects, upload and

retrain content, read conversations, manage contacts, pull analytics — use Platform MCP.

Available on Professional, Advanced, Agency and Enterprise plans.

Get your config

1. Go to Organization Settings → API Tokens

2. Create an API token (it starts with api-v1-), or copy an existing one

3. Click MCP at the top of the page and pick your tool

Configuration

{
  "mcpServers": {
    "yourgpt_org_mcp": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.yourgpt.ai/v1/platform/mcp",
        "--header",
        "Authorization: Bearer api-v1-your_api_token"
      ]
    }
  }
}

Replace api-v1-your_api_token with your real token. Config file locations are the

same as above — Claude Desktop, Cursor, Windsurf, and any other MCP client.

What your assistant can do

For example, you can ask your assistant to:

• "Try this support persona in the Agent Playground across a few AI models, then apply the best one to my chatbot."

• "My chatbot gave a wrong answer this morning — show me how it worked out that reply and what it pulled from my knowledgebase."

• "Add these 12 pages to my chatbot's training and tell me when they're ready."

Your token's permissions and project scope apply to MCP exactly as they do to the

REST API — an assistant can only do what the token is allowed to do.

Full endpoint reference: https://docs.yourgpt.ai/chatbot/api-reference/

Additional Notes

• Keep your token private — anyone who has it gets the same access you granted.

• Chatbot Knowledgebase MCP: each chatbot has its own setup. You can remove or regenerate its config from the dashboard at any time.

• Platform MCP: one setup per organization. To cut off access immediately, disable or revoke the token under Organization Settings → API Tokens.

If you have any questions about the integration, feel free to join our Discord community — our technical team is there and ready to help you directly.

The Model Context Protocol (MCP) lets you connect your YourGPT chatbot as a knowledgebase inside tools like Claude Desktop, Cursor, and Windsurf.

This guide help you through two main parts:

1. Setting up MCP inside YourGPT

2. Connecting it with your tool (Claude Desktop, Cursor, or Windsurf)

1. Configure MCP in YourGPT

Before connecting with any tool, you must activate the MCP server from your YourGPT dashboard.

Step 1: Open the MCP Integration

1. Go to YourGPT Dashboard:

2. Then Integrations → MCP (Model Context Protocol)

3. Click Configure.

Step 2: Create MCP Server

Click on Create MCP to generate your chatbot’s MCP configuration.

Step 3: Select Your Target Tool

After the server is created, you’ll see a configuration screen with tool-specific tabs for:

• Claude Desktop

• Cursor

• Windsurf

Each tab gives you a JSON block to copy, along with instructions on where to place it.

Part 2: Connect MCP with Your Tool

Now that your chatbot is MCP-ready, here’s how to connect it with each tool.

2.1 Claude Desktop

Step 1: Open Claude Settings

Click Claude in your menu bar and select Settings.

Step 2: Enable Developer Mode

From the settings sidebar, go to Developer and click Edit Config.

Step 3: Locate Configuration File

On macOS, go to:

~/Library/Application Support/Claude/claude_desktop_config.json

If it’s not there, relaunch Claude or check permissions.

Step 4: Add Your MCP Configuration

Open the config file in any text editor or IDE, and paste the code provided in YourGPT.

Paste the configuration block provided by YourGPT. It should look like this:

{
  "mcpServers": {
    "yourgpt_mcp": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.yourgpt.ai/sse",
        "--header",
        "Authorization: Bearer YOUR_KEY"
      ]
    }
  }
}

Step 5: Restart Claude and Confirm

Save your changes, Restart Claude Desktop.

If MCP is working correctly, you’ll see a hammer icon in the bottom-right corner.

2.2 Cursor Setup

Cursor setup is much simpler.

1. Look for Cursor Settings, then click on MCP, the add new global MCP server

2. Locate or create the file:

~/.cursor/mcp.json

3. Paste your MCP config JSON into this file.

4. Save and restart Cursor.

5. Test Your MCP.

2.3 Windsurf Setup

For Windsurf, follow the same approach.

Step 1: Locate the MCP Config File

~/.codeium/windsurf/mcp_config.json

Step 2: Paste Your Configuration

Use the config you got from YourGPT. It should look like this:

{
  "mcpServers": {
    "yourgpt_mcp": {
        "command":"npx",
        "args": [
            "mcp-remote",
            "https://mcp.yourgpt.ai/sse",
            "--header",
            "Authorization: Bearer YourKEY"
        ]
    }
  }
}

Paste it into mcp_config.json using your editor.

Step 3: Refresh Windsurf MCP & Test

Open Windsurf and look for the MCP section in the right panel. Click on Refresh.

Once it connects, your AI will appear under “available MCP servers”.

Additional Notes

• Keep your Bearer token private.

• You can remove and regenerate your MCP config from the YourGPT dashboard at any time.

• Each chatbot in YourGPT can have its own MCP setup.

If you have any questions about the integration, feel free to join our Discord community — our technical team is there and ready to help you directly.