Run scraping robots on a fixed schedule

Browse AI runs pre-built robots that extract structured data, such as a table of listings or prices, from a specific page layout. A Beam agent triggers a robot on a schedule or on demand, reads the extracted rows once the run completes, and applies an approved rule, such as filtering out rows that already exist or matching new rows against an internal list. Runs that return no data, far fewer rows than expected, or an error suggesting the page layout changed are routed to a person to check the robot's setup.

Detect and report changes on a monitored page

A core use of Browse AI is watching a page for a change, such as a price drop or a new listing appearing. A Beam agent reads the comparison between the current and previous extraction and applies an approved rule when a change is detected, such as updating a tracked value or sending a notification. This catches changes without someone checking the page manually. Where a detected change looks unusually large or the page structure appears to have shifted rather than the actual content, the case is flagged for a person to confirm.

Feed extracted rows into another system

Once a Browse AI robot finishes a run, the extracted rows typically need to reach another system, such as a spreadsheet or database, to be useful. A Beam agent reads the completed run's output and writes new or changed rows to the approved destination, applying rules like deduplication or field mapping along the way. Rows that don't map cleanly to the expected fields, or a run that returns data in an unexpected format, are held for a person to review before they are passed downstream.