jupyter-live-kernel

Use a live Jupyter kernel via the jupyter_live_kernel.py script's execute command, which keeps variables, imports, and objects alive across calls. Point it at a notebook path on a running Jupyter server and each execution builds on the previous state.

Use a live kernel for iterative exploration, data science, and ML tasks where you build state incrementally. Use one-shot execution for stateless scripts that need tool access, and the terminal for shell commands, installs, and process management.

You need uv installed, JupyterLab installed via uv tool install, and a running Jupyter server. A kernel session must also be started through the Jupyter REST API before the execute command can run code.

The kernel needs a moment to initialize after the server starts, so the first execution may hit the 30-second default timeout. Retrying usually succeeds, and you can pass a longer timeout such as --timeout 120 for heavy operations.

Yes, the variables subcommand lists all live variables and can preview a specific one by name. This lets you check intermediate DataFrames or objects without re-executing the code that created them.