jupyter-mcp-server

What it is and what it does

Jupyter MCP Server is a Model Context Protocol server that bridges Jupyter notebooks and AI agents, allowing tools to read, write, and execute code in notebooks in real-time. It exposes MCP tools for notebook management (switching between notebooks, listing kernels), cell operations (reading and executing cells, managing outputs), and code sandbox execution across multiple backends including local Jupyter, Datalayer, Kaggle, Google Colab, and Modal.

The package is built on fastapi and jupyter-server, with observability via opentelemetry-api and opentelemetry-sdk. It can run as a standalone MCP server, as a Jupyter extension, or connect to existing Jupyter deployments. Datalayer hosts a public instance for immediate use without local setup. The primary use case is enabling AI agents to work with notebooks persistently—executing code, handling failures, and switching execution backends on demand.

Use it for:

• Enable AI agents to execute code in local or cloud Jupyter notebooks with real-time feedback and error recovery.
• Route code execution across multiple sandbox backends without changing client code.
• Build multi-notebook workflows where an agent switches between notebooks and manages kernel state automatically.
• Integrate Jupyter execution into an MCP-compatible agent framework for persistent, context-aware data analysis.
• Use the hosted Datalayer MCP endpoint to connect any MCP client to notebooks without running a local server.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

An MCP server that exposes Jupyter notebooks and code execution to AI agents and clients, enabling real-time notebook management, cell execution, and sandbox-backed code runs across local and cloud environments.

Yes, if you use Jupyter with AI agents or need to expose notebook execution via MCP. The package is actively maintained, permissively licensed, has low install friction, and solves a specific gap: real-time agent control of notebooks. Install it locally to connect your Jupyter server, or use the hosted endpoint for immediate integration. No known vulnerabilities as of 2026-08-14.

Install

jupyter-mcp-server on PyPI

pip install jupyter-mcp-server

uv add jupyter-mcp-server

poetry add jupyter-mcp-server

Installing jupyter-mcp-server

Before you install

Low install friction with a pure-Python wheel. Active maintenance (last commit 2026-08-14, 1244 stars) and recent release cycle. Requires Python >=3.10 and 15 runtime dependencies including fastapi, jupyter-server, mcp, and opentelemetry libraries—a substantial but standard stack for Jupyter tooling.

License in practice

BSD 3-Clause License (permissive) allows commercial use, modification, and redistribution with minimal restrictions—only requiring copyright notice and disclaimer retention.

Quickstart

pip install jupyter-mcp-server

from jupyter_mcp_server import JupyterMCPServer
import asyncio

server = JupyterMCPServer()
await server.connect_to_jupyter()

Requires Python >=3.10 and a running Jupyter server instance to connect to; MCP client needed to interact with the server.

Verify before relying

• Whether the package can run standalone or only as a Jupyter extension or external server process.
• Performance characteristics under concurrent notebook and sandbox operations.
• Compatibility matrix with specific Jupyter Server and MCP client versions.
• Whether GPU sandbox variants require additional credentials or setup beyond the package itself.

Package facts

Evidence: jupyter_mcp_server-1.3.8-py3-none-any.whl

Keywords: Jupyter

Tags

More Application Frameworks packages

Further reading