page-agent in-page GUI agent · trending #10
alibaba / page-agent

Say it. The page does it.

A JavaScript in-page GUI agent — control any web interface with natural language. It reads the live DOM, finds the elements, and acts.

JS · drop-in reads the live DOM no selectors
AGENT // PAGE CONTROLDOM LINKED
>"filter to unread, then archive the newsletters"
01LOCATEFilter control · button[aria-label=filter]found
02CLICKUnread · menuitem "Unread"done
03SELECTRows matching "newsletter" · 6 of 18done
04ARCHIVESelected rows · toolbar actionacting
natural language → DOM actionsillustrative run
Open source · works on interfaces you don't control github.com/alibaba/page-agent ↗