What It Does
Page Understanding reduces a large, noisy live page into a structured state object. It is designed for decision layers that need to know what is on the page without sending a full raw DOM dump.
Toolkit
Page Understanding
Builds a compact structured snapshot of the page so another tool or workflow can make better decisions with less noise.
How It Works
- Captures a snapshot grouped into major regions such as header, body, and footer.
- Adds compact summaries of visible sections, links, and controls.
- Separates regular inputs from upload fields.
- Builds a form summary with counts for required, filled, empty, and upload inputs.
- Returns the result as structured JSON for downstream automation.
Best For
- Decision layers that need current page state
- Multi-step automations
- Cases where a raw DOM or full HTML payload would be too noisy or expensive
Behavior Notes
- It is not meant to directly change the page.
- It is designed to reduce token usage by sending a compact state object instead of raw page content.