Notebook Lab

The lesson runtime. It opens a lesson notebook, runs its cells against a Python kernel, and runs its agent tasks against an MCP server you wrote. On its own it teaches nothing, because the Learn extension supplies the lessons' gates, hints and tutor.

What you can do

• Open a lesson notebook and work through it cell by cell.
• Run cells against the kernel you pick, with output that streams in as it runs.
• Run the whole lesson at once, stopping where it asks for your prediction.
• Watch a real agent decide whether to call the tool you wrote, step by step.
• Ask the tutor about the cell you are stuck on.
• Switch to teaching mode to write a lesson rather than take it.

Getting started

Double-click a lesson file (its name ends in .lesson.ipynb), or pick a lesson in the Learn window. A plain .ipynb still opens in the regular notebook editor.

Requirements

• Learn. It plans every lesson: which cells are gated, which hints exist, and what your tutor says. Without it a lesson does not open, because a lesson with its gates silently switched off is not a lesson.
• A Python kernel, for lessons that run code.
• For lessons that test an agent, your MCP server connected from the MCP panel, and a model added from the LLM Capabilities panel.

Good to know

• An agent lesson connects to exactly one MCP server. If more than one is connected, it asks you to disconnect the others rather than guess which is yours.
• Progress is single-user. It belongs to whoever has the workspace open, because there is no learner identity yet.