jupyter-mcp-server
View on GitHub🪐 🔧 Model Context Protocol (MCP) Server for Jupyter.
MCP server that lets AI agents connect to and control Jupyter Notebooks in real time: browse files, manage notebooks/kernels, edit and run cells, and read multimodal outputs. Works with any MCP client and can execute code in local or cloud code sandboxes.
Use Cases
Let AI agents read, edit and execute Jupyter notebook cells in real timeRun code in remote or cloud sandboxes (Datalayer, Colab, Kaggle, Modal, E2B, Daytona)Connect Claude Code / Claude Desktop / Cursor to a live Jupyter Server via MCPGive agents notebook context with multimodal cell outputs (images, plots, text)Automate notebook creation, kernel restarts and cell reordering from an agentCite specific notebook cells as prompt context via MCP promptsTrace agent tool calls and kernel executions with OpenTelemetryRun JupyterLab commands as MCP tools through jupyter-mcp-tools
Built With
- Language
- Python
- Frameworks
- MCP Python SDK · FastAPI · Typer · Uvicorn · Pydantic · OpenTelemetry · Jupyter Server · JupyterLab · jupyter-collaboration · pycrdt · Docker
Tags
mcp · mcp-server · jupyter · notebooks · code-execution · jupyterlab · code-sandboxes · python · oauth · opentelemetry · multimodal · jupyterhub · claude-code · streamable-http