Tutorial: Run Jupyter MCP Server with Docker in 5 Minutes

In this tutorial you'll get Jupyter MCP Server running with Docker and connect it to an MCP-compatible AI assistant — in under five minutes. We'll use the official image, verify the server responds, and run your first code cell through the agent.

Step 1 — Pull and run the container

Create a project directory and start the server. The image listens on port 8080 by default and keeps notebook data under ./data:

docker run -d --name jupyter-mcp \
  -p 8080:8080 \
  -v $(pwd)/data:/data \
  datalayer/jupyter-mcp-server:latest

Step 2 — Verify it's alive

Check the health endpoint. A 200 response means the MCP server is ready to accept tool calls:

curl -s -o /dev/null -w "%{http_code}\n" http://localhost:8080

Step 3 — Connect your AI assistant

Point any MCP client at the server URL. Your assistant will now discover tools for managing notebooks and kernels, and you can ask it to:

• Create a new notebook session
• Execute a cell that computes a data summary
• Read back the output and iterate on the analysis

Step 4 — Keep it running like a pro

For production, add restart: unless-stopped and mount a persistent volume so your notebooks survive container restarts. A complete docker-compose.yml is generated for you on the tool page.

That's it — your AI agent now has a live Jupyter backend. Experiment with multi-cell workflows and let the agent handle the plumbing while you focus on the analysis.