Core Concepts
Understand notebooks, blocks, integrations, and how they work together.
The Notebook
A notebook is a workspace for building multi-step workflows. Think of it as a project container — everything related to a specific workflow lives here: the apps it connects to, the instructions you write, and the outputs it generates.
Each notebook operates independently. You might have one notebook for weekly reporting, another for customer onboarding emails, and a third for competitive research. They don't share state or context unless you explicitly copy content between them.
At the top of every notebook is the Integrations panel. This controls which of your connected apps are available to blocks within that notebook. Enable only what you need — a notebook for calendar management doesn't require access to your CRM.
Blocks
Blocks are the building units of a notebook. Each block contains an instruction (your prompt), an output (the result), and controls for execution.
Writing Instructions
In the input area, describe what you want in natural language. Be specific about sources, filters, and desired output:
- •"Pull all Slack messages from #product-updates in the last week"
- •"Find emails from investors that mention 'term sheet'"
- •"Summarize the three most recent Linear tickets assigned to me"
Running a Block
Click Run to execute. The block status changes from "Ready" to "Running," and you'll see the AI's reasoning process as it works. When complete, the output appears below your instructions.
Outputs and Artifacts
Every block produces an output. This might be text, a summary, a list, or a structured artifact like a Google Doc or spreadsheet. You control the format by specifying it in your instructions:
- •"...and format the results as a markdown table"
- •"...create a Google Doc with this content"
- •"...output as a CSV I can import into Airtable"
Outputs render directly in the notebook. For file-based artifacts, Deep Notebook creates the actual file in your connected app — a real Google Doc in your Drive, not just a preview.
Chaining Blocks
Blocks execute in sequence. Each block can reference outputs from previous blocks, creating workflows that span multiple steps and apps.
This chaining is what transforms simple queries into sophisticated automations. The AI automatically understands context from previous blocks.
The Magic Prompt
Every block includes a Refine Prompt button (✨) that helps you write better instructions. Click it, and Deep Notebook analyzes your prompt and suggests improvements.
The magic prompt helps with:
- •Clarifying ambiguous references
- •Adding useful parameters you might have missed
- •Restructuring for better execution
- •Suggesting output formats
Block States
Blocks display their current status to help you track execution progress:
Attachments
Use the paperclip icon (📎) to attach files directly to a block. These become context for your instructions — attach a PDF and ask for a summary, or include a spreadsheet and request analysis. The AI processes the attachment as part of the block's execution.
Run All
When you have multiple blocks, click Run All to execute them in sequence. This is useful for running an entire workflow from start to finish, or re-running after making changes to earlier blocks.