pykernel-mcp 0.2.5

MCP server providing persistent IPython kernel for executing Python code with numpy, pandas, and matplotlib

PyKernel MCP

MCP server to make it possible for an agent to execute python in a Jupyter kernel.

Features

PyKernel provides a persistent IPython kernel environment for executing Python code through the Model Context Protocol. After setting this server up, your agent will be able to:

• Maintains state between executions - variables, imports, and functions persist across tool calls
• Pre-loaded scientific stack - comes with numpy, pandas, and matplotlib already imported
• Rich output support - captures text output, errors, and matplotlib plots
• Visualizations - inline matplotlib plots rendered as images
• Package installation - install additional packages on-the-fly with the install_package tool
• Kernel management - restart the kernel to clear state when needed

Use Cases

• Quick data analysis and exploration without writing files
• Iterative computation where you build on previous results
• Mathematical calculations and statistical analysis
• Data visualization with matplotlib
• Testing Python code snippets
• Prototyping algorithms with maintained state

The kernel automatically handles execution timeouts, captures both stdout and stderr, and provides detailed error tracebacks when code fails.

Test

Just execute:

npx @modelcontextprotocol/inspector uv run src/pykernel_mcp/server.py

Installation

Click the button to install:

Or install manually:

Go to Advanced settings -> Extensions -> Add custom extension. Name to your liking, use type STDIO, and set the command to uvx pykernel-mcp. Click "Add Extension".