Skip to main content

Getting Started

To use Jupyter MCP Server, you first need to decide which setup fits your needs:

  • MCP Server Location - It can be running as a Standalone MCP Server or inside the Jupyter Server as a Jupyter Server Extension - The last one has for advantage to avoid running 2 separate servers (Jupyter server + MCP server) but only supports Streamable HTTP transport.
  • MCP Transport - It supports both STDIO Transport and Streamable HTTP Transport
  • Jupyter Provider - It can connect to a local or remote Jupyter server, to JupyterHub, to Datalayer or to Google Colab hosted Notebooks.

This guide will help you set up a Jupyter MCP Server to connect your preferred MCP client to a JupyterLab instance.

The Jupyter MCP Server acts as a bridge between the MCP client and the JupyterLab server, allowing you to interact with Jupyter notebooks seamlessly.

You can customize the setup further based on your requirements. Refer to the server configuration for more details on the possible configurations.

Choosing Your Transport​

Jupyter MCP Server supports two types of transport to connect to your MCP client: STDIO and Streamable HTTP. Choose the one that best fits your needs.

For more details on the different transports, refer to the official MCP documentation here.

STDIO Transport​

Best for: Desktop applications, Docker deployments, single-user setups

  • ✅ Simple configuration
  • ✅ Works with most MCP clients (Claude Desktop, Cursor, Windsurf, VS Code)
  • ✅ No additional server ports needed
  • ❌ One client connection at a time

👉 Get started with STDIO Transport

Streamable HTTP Transport​

Best for: Web applications, multiple concurrent clients, production deployments

  • ✅ Multiple clients can connect simultaneously
  • ✅ Web-based access
  • ✅ Can run as Jupyter Server Extension (no separate MCP server process)
  • ❌ Requires opening network ports

If you choose Streamable HTTP transport, you can also choose to run the MCP server:

Provider Options​

Local/Remote Jupyter Server​

Connect to any standard JupyterLab or Jupyter Notebook server.

👉 Choose your transport:

JupyterHub​

Multi-user deployments with authentication and resource management.

👉 JupyterHub Setup Guide

Multi-User Deployments

If you're deploying for multiple users, see the Multi-User Documentation for architecture patterns and best practices.

Datalayer Platform​

Enterprise-grade hosted Jupyter with built-in collaboration and security features.

👉 Datalayer Provider Documentation

Google Colab​

Connect to Google Colab notebooks (experimental).

👉 Google Colab Provider Documentation

MCP Client Configuration​

Once you've set up your MCP server, you need to configure your MCP client. Choose your client:

Next Steps​

After setting up your MCP server and client:

🔧 Explore Available Tools - Learn about the MCP tools for interacting with notebooks

📝 Use Prompts - Discover prompt features for citing and referencing notebooks

🔒 Secure Your Deployment - Review security best practices for token management and authentication

📚 Learn More - Check out the Resources section for tutorials, videos, and community content

Additional Resources​