Claude Code Jupyter

IPython/Jupyter magic commands for interacting with Claude. This extension allows Claude to generate code that executes in your current Jupyter notebook or IPython session, sharing all variables and state.

Installation

pip install claude-code-jupyter

For development installation:

git clone https://github.com/anthropics/claude-code-jupyter.git
cd claude-code-jupyter
pip install -e ".[dev]"

Quick Start

In a Jupyter Notebook

1. Load the extension:

%load_ext cc_jupyter

2. Use Claude to generate and execute code:

%cc Create a list of prime numbers up to 100

3. To see available options and usage information:

%cc --help

In IPython

1. Start IPython:

ipython

2. Load the magic:

%load_ext cc_jupyter

3. Use Claude to generate and execute code:

%cc Create a list of prime numbers up to 100

When to use %cc vs %%cc

• Use %cc (single %) for:

  • Short, one-line instructions
  • Command options like --help or --verbose
  • Continuing without additional instructions

• Use %%cc (double %) for:

  • Multi-line instructions or detailed prompts
  • Complex requests that need more explanation
  • Any prompt ending with ? (to avoid IPython's help system)

Features

Automatic Code Writing

Claude's code blocks are automatically extracted and written into a new cell that you approve by executing:

import numpy as np
data = np.random.randn(1000)

%%cc
Create a histogram of the data array and calculate its statistics

Continue Conversations

Follow up on previous requests:

%cc Create a pandas DataFrame with sales data

# Claude writes code into a new cell, you review it and execute it

%cc Now add a column for profit margin

Shared State

All variables are shared between you and Claude:

# You create:
x = [1, 2, 3, 4, 5]

# Claude can use:
%cc Square each element in x and store in a new list called x_squared

# Claude writes code into a new cell, you review it and execute it

# You can use:
print(x_squared)  # Claude's variable is available!

How It Works

1. You give Claude a task via %cc <task>
2. Claude generates Python code based on your request
3. The extension creates a new cell with Claude's Python code
4. You review Claude's code and execute it
5. You can continue working with Claude via %cc

Examples

Data Analysis

%%cc
Load the iris dataset from sklearn and create a scatter plot matrix

Interactive Development

# Your code
data = load_my_data()

%cc Clean the data by removing NaN values and outliers

# Claude writes code into a new cell, you review it and execute it

%cc Now normalize the numerical columns

# data is now cleaned and normalized in your session

Development

Setup

# Clone the repository
git clone https://github.com/anthropics/claude-code-jupyter.git
cd claude-code-jupyter

# Install in development mode with dev dependencies
pip install -e ".[dev]"

# Install pre-commit hooks
pre-commit install

Running Tests

pytest

Code Quality

# Run linting
ruff check .

# Run formatting
ruff format .

# Run type checking
pyright

Safety

• Claude does not execute code by itself. All code is reviewed by a human, who can choose to execute (or not).

Limitations

• Claude can write code with errors
• No sandboxing - code runs with same access as rest of notebook

License

MIT License - see LICENSE for details.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

This extension is perfect for interactive data analysis, learning, and development where you want Claude to work alongside you in the same environment!