Jupyter Notebooks

Jupyter Notebooks skill connects your Claude agent to a running Jupyter server, enabling notebook creation, cell execution, kernel management, and output retrieval — all via the Jupyter REST API. Your agent can open existing notebooks, execute specific cells or entire notebooks, read cell outputs (including rich outputs like plots and DataFrames), and save results.

Notebook creation from scratch is fully supported: your agent generates a new notebook with properly structured cells — Markdown explanations, Python code, and structured outputs — creating a complete, documented analysis workflow that humans can review and extend in the Jupyter UI. This bridges AI-generated analysis with the interactive exploration model data scientists rely on.

Kernel management handles the full lifecycle: start kernels, interrupt hung computations, restart after crashes, and shut down when done. Variable inspection lets the agent query the current kernel's namespace — 'what variables are defined in this notebook?' — useful for understanding the state of long-running analyses without re-executing all cells.

JupyterHub multi-user environments are supported via token authentication, enabling integration with team notebook servers. The skill also reads notebook metadata and execution counts, making it useful for auditing which notebooks have stale outputs (last executed months ago) versus actively maintained analyses. Data scientists, ML engineers, and quantitative researchers use Jupyter Notebooks skill to build AI-assisted analysis pipelines that live natively in their existing notebook environment.

Installation

clawhub install jupyter

Install: clawhub install jupyter