Deepnote Agent

We’re introducing Deepnote Agent, the next evolution of AI collaborators for data notebooks. Agent offers a unified chat experience that turns your notebook into a flexible canvas. You can ask for anything from simple edits like ‘refactor variable df’ to complex tasks like ‘make this notebook presentation-ready’.

Agent interprets your intent, creates a clear plan, and executes it step by step. It can create, edit, and remove SQL, Python, or text blocks — anywhere in your notebook — while staying fully aware of the project context.

Accessing Deepnote AI

If you’re on a paid plan or Team trial, open Agent from:

• the Agent button in the top bar of your project, or
• the floating Agent button in the bottom-right corner.

The chat window will open on the side. Expand it into a full sidebar for longer conversations, or collapse it to focus on your notebook.

Using Agent

Agent has two main modes:

• Edit mode (default) — Agent makes direct changes to your notebook. It can add, edit, or delete content; execute code blocks; inspect outputs; and adapt its actions based on context.
• Ask mode — Agent won’t edit anything. Use it to brainstorm next steps, ask questions about your data, or get help with Deepnote features.

In Edit mode, you can request anything from straightforward edits (adding analysis, fixing a SQL query, editing code) to large, notebook-wide changes (refactoring variables, cleaning up blocks, fixing multiple errors, or adding documentation).

No matter the scope, Agent always:

1. Interprets your request.
2. Creates a transparent, step-by-step plan.
3. Executes each step systematically.

While running, Agent shows all changes in real time. You can click any action item to jump to the relevant block.

When a run finishes, you’ll see a summary of changes with a before-and-after diff view for code edits. You can undo all changes from a run by clicking the bin icon. If you follow up with a new request, Agent continues from the updated state of your notebook.

Leaving feedback

Deepnote Agent is in Beta, and we’re improving it continuously. New functionality will be added throughout the Beta period, so stay tuned for updates.

We’d love your feedback as we shape the full release:

• If you downvote an AI response, a pop-up will open where you can add comments and share feedback directly with the team.
• We’ll also share a short survey with the most active participants — your insights will help us prioritize improvements.

Your input is invaluable in helping us refine Agent into the best possible collaborator for your notebooks!