Skip to main content

Installation

Detailed installation instructions for MCP Jupyter Server.

Prerequisites​

1. Install UV​

UV is required for running MCP Jupyter. Install it using one of these methods:

# macOS/Linux
curl -LsSf https://astral.sh/uv/install.sh | sh

# Windows
powershell -c "irm https://astral.sh/uv/install.ps1 | iex"

# Or via pip
pip install uv

2. JupyterLab Setup​

MCP Jupyter requires a running JupyterLab server with specific extensions:

# Create virtual environment
uv venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

# Install required packages
uv pip install jupyterlab jupyter-collaboration ipykernel

# Optional: Install additional packages
uv pip install numpy pandas matplotlib

Server Installation​

The simplest way to use MCP Jupyter is via uvx:

uvx mcp-jupyter

This command will automatically download and run the latest version.

Transport Modes​

MCP Jupyter supports two transport protocols:

  • stdio (default) - Standard input/output communication, ideal for IDE integration
  • http - Streamable HTTP transport with session management, ideal for web clients and remote access

Use Cases for HTTP Transport​

  • Serverless deployments: Host the MCP server in cloud environments (AWS Lambda, Google Cloud Functions, etc.)
  • Remote access: Connect to the server from different machines or networks
  • Web integrations: Build web-based AI assistants that connect to the MCP server
  • Multi-user environments: Deploy a central MCP server that multiple clients can connect to
  • Stateless operations: Use --stateless-http for environments where session persistence isn't needed or desired

Using HTTP Transport​

Start the server with HTTP transport:

# HTTP transport on default port 8000
uvx mcp-jupyter --transport http

# HTTP transport on custom port
uvx mcp-jupyter --transport http --port 8090

# HTTP transport in stateless mode (no session persistence)
uvx mcp-jupyter --transport http --port 8090 --stateless-http

From Source​

For development or customization:

# Clone the repository
git clone https://github.com/block/mcp-jupyter.git
cd mcp-jupyter

# Sync all dependencies including dev tools
uv sync

Starting the Jupyter Server​

Basic Setup​

jupyter lab --port 8888 --IdentityProvider.token BLOCK --ip 0.0.0.0

Custom Configuration​

For production use or specific setups:

# Set custom token
export TOKEN=your-secure-token
jupyter lab --port 8888 --IdentityProvider.token $TOKEN

# Use config file
jupyter lab --config=/path/to/jupyter_config.py

Client Configuration​

Goose​

Add to your Goose session:

goose session --with-extension "uvx mcp-jupyter"

Or for development:

goose session --with-extension "uv run /path/to/mcp-jupyter/.venv/bin/mcp-jupyter"

Cursor​

Add to your .cursor/mcp.json:

{
  "mcpServers": {
    "jupyter": {
      "command": "uvx",
      "args": ["mcp-jupyter"],
      "env": {
        "TOKEN": "your-token-here"
      }
    }
  }
}

Option 2: HTTP Transport​

For HTTP transport, first start the server separately:

uvx mcp-jupyter --transport http --port 8090

Then configure Cursor's .cursor/mcp.json:

{
  "mcpServers": {
    "jupyter-http": {
      "url": "http://localhost:8090/mcp/"  // ⚠️ Trailing slash is REQUIRED
    }
  }
}
Important

The trailing slash (/mcp/) is required for Cursor to connect properly to the HTTP endpoint.

Other MCP Clients​

The general pattern for any MCP client:

{
  "command": "uvx",
  "args": ["mcp-jupyter"],
  "stdio": true
}

Troubleshooting​

Common Issues​

  1. "Jupyter server is not accessible"

    • Ensure Jupyter is running on the expected port
    • Check firewall settings
    • Verify the token matches

  2. "No kernel found"

    • Make sure you have opened a notebook in Jupyter
    • Check that ipykernel is installed
    • Verify the notebook path is correct

  3. Package installation fails

    • Ensure your virtual environment has pip
    • Check write permissions
    • Verify internet connectivity

Debug Mode​

Enable debug logging:

export MCP_JUPYTER_DEBUG=1
uvx mcp-jupyter

Next Steps​