notebook-tools

name	notebook-tools
description	Use these tools whenever the user asks to read, modify, or clean up Jupyter notebooks (.ipynb files).

Working with Jupyter Notebooks

Workflow

1. Create with notebook_create. Do not hand-write notebook JSON.
2. Start by listing or searching. Call notebook_list_cells for an overview or notebook_search when the target cell is known by text. Both return stable cell ids.
3. Read targeted cells. If you need source content, call notebook_read with cell_ids, cell_range, or include_outputs only when needed.
4. Address by cell_id, not index. After any insert/delete, indexes shift. Cell ids are stable across operations.
5. Use stale guards when available. Pass expected_source_sha256 or expected_file_mtime_ms from the latest read/list/search result before mutating.
6. Edit, don't rewrite. Use notebook_edit_cell for source changes, notebook_insert_cell to add cells, and notebook_validate after changes.

After mutations

• Editing a code cell automatically clears its outputs. Mention this in your reply.
• Tell the user the new cell id when you insert.
• If the user asked for a "clean notebook", run notebook_clear_outputs with no cell_id to clear all outputs at once.
• Run notebook_validate before claiming that a changed notebook is valid.

Don't

• Don't use read / write / edit core tools on .ipynb files.
• Don't try to execute cells — that's not supported. Suggest the user run the notebook in JupyterLab themselves.
• Don't address cells by index across multiple tool calls — always re-list or use ids.