Connect MCP clients - Squirrels Docs

Every Squirrels project has an MCP server running at /mcp automatically. Thus, AI tools that support the Model Context Protocol can be used to ask questions about the datasets configured in a Squirrels project. This guide assumes that you have a Squirrels server running at http://localhost:8000 in standalone mode.

Authorization header (when required)

If your project uses auth_type: required, your MCP client must authenticate.

For auth_strategy: external, MCP requests must include Authorization: Bearer <token> where <token> is a provider-issued OAuth access token (or JWT).
- If authentication is missing/invalid, the server responds with HTTP 401 and a WWW-Authenticate header that includes a resource_metadata URL (usually served at /.well-known/oauth-protected-resource).
- MCP clients that support OAuth discovery can use this to find the correct authorization server to authenticate with and fetch the access token from first.
For auth_strategy: managed, MCP requests can include x-api-key: <api_key> as a header where <api_key> is an API key issued by the Squirrels API server.

MCP clients

Choose which AI tool you want to use to interact with your Squirrels project.

Claude Desktop
Claude Code
Cursor
VS Code / GitHub Copilot

Prerequisites:

Claude Desktop installed
Node.js (>= v18.0.0) and npm: Download from nodejs.org
A Claude.ai account (the free plan works!)

Steps:

Open Claude Desktop.
Click your name in the lower left corner, then click Settings.
Click Developer under “Desktop app”.
Click the Edit Config button. This opens the file location of claude_desktop_config.json.
Open claude_desktop_config.json in a text editor, and paste in the following:

{
  "mcpServers": {
    "squirrels-analytics": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote", 
        "http://localhost:8000/mcp"
      ]
    }
  }
}

Save the file and exit Claude Desktop (this is different than closing the window).

To exit Claude Desktop on Windows, click the menu icon on the top left corner and then click File -> Exit

Open Claude Desktop again. You should now see the Squirrels project enabled by clicking the “search and tools” icon below the chat input.
Ask questions about the datasets in the Squirrels project. See the section below for example prompts.

Prerequisites:

Claude Code (CLI) installed
Either:
- A Claude.ai account with Pro plan or higher, or
- A Claude Console account with a credit balance

Steps:

Add the MCP server to Claude Code with name “squirrels-analytics” and URL http://localhost:8000/mcp.

claude mcp add --transport http squirrels-analytics http://localhost:8000/mcp

Use the following commands to confirm details about the MCP server.

# List all configured servers
claude mcp list

# Get details about the MCP server you just added
claude mcp get squirrels-analytics

Start Claude Code by running claude in the terminal. You can check the server status by typing /mcp in the chat input.

# Start Claude Code
claude

# Check server status (within Claude Code)
/mcp

Ask questions about the datasets in the Squirrels project. See the section below for example prompts.

Prerequisites:

Cursor (IDE) installed
A Cursor account

Steps:

Open Cursor.
Go to File -> Preferences -> Cursor Settings.
Click on Tools & MCP in the left panel.
Click the Add Custom MCP button. This should open a mcp.json file.
In the mcp.json file, paste the following and save.

{
  "mcpServers": {
    "squirrels-analytics": {
      "url": "http://localhost:8000/mcp",
      "headers": {}
    }
  }
}

Press Ctrl+I to open the AI agent chat interface. Use Ctrl+T if you need to start a new chat.
Ask questions about the datasets in the Squirrels project. See the section below for example prompts.

Prerequisites:

VS Code (IDE) installed
VS Code extensions: GitHub Copilot and GitHub Copilot Chat installed
A GitHub account

Steps:

Open VS Code.
Create a .vscode folder in your project’s root directory if it doesn’t exist.
Create and open the mcp.json file in the .vscode folder.
In the mcp.json file, paste the following and save.

{
  "servers": {
    "squirrels-analytics": {
      "url": "http://localhost:8000/mcp",
      "headers": {}
    }
  }
}

Click Start on the MCP server.
Press Ctrl+Shift+I to open the AI agent chat interface. Use Ctrl+N if you need to start a new chat.
Ask questions about the datasets in the Squirrels project. See the section below for example prompts.

Example prompts

The following are sample prompts that can be used for any Squirrels project when interacting with an AI chat agent:

What datasets do you have access to?

What tools do you have access to?

Other prompts to ask depend on the datasets in your Squirrels project.

AI

Python integration

​Authorization header (when required)

​MCP clients

​Example prompts

Authorization header (when required)

MCP clients

Example prompts