[Feature Request] Enable Agents to Automatically Construct Browser Automation Flows Using the Browser Toolkit #3477
Description
Required prerequisites
- I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
- Consider asking first in a Discussion.
Motivation
Many browser-related tasks require building fast, fully reusable automation flows where the only changing parts are input variables, while the underlying sequence of steps remains fixed. Traditional web-scraping scripts are slow to develop, and letting the agent drive the browser live on every run can be unstable.
We can leverage the agent’s capabilities to automatically construct these reusable flows. While performing the task once, the agent can record all relevant page elements and their exact positions, then use this historical information to generate a stable automation script.
However, to make such flows dependable, we need a robust and stable element-fetching mechanism. Conventional selectors—such as basic XPath expressions or raw text labels—are often ambiguous or prone to change, which makes the generated flows brittle and not reliably reusable.
Solution
No response
Alternatives
No response
Additional context
No response