Summary
Clicking by visible text is the difference between a script that survives a redesign and one that does not.
Alongside the usual selector-based interaction it can act on the words a person would actually see, which is how an agent describes what it wants — and reading the page back as text rather than as a screenshot keeps the result usable.
What it is
A Playwright-backed browser server: it opens pages, interacts with them by CSS selector or by visible text, and returns what is on screen as text, HTML or an image.
What you get
- Navigation to a URL, which starts a session automatically if none is open
- Clicks by CSS selector, or by the text that is actually visible on the element
- Input fields filled by selector and value
- Screenshots of the whole page or of one element
- The visible text of the page, or the HTML under a selector you name
- Arbitrary JavaScript executed in the page console
Requirements
Python with Playwright available, so the browser it drives has to be installed on the machine running the server.
Setup effort
One command — uvx playwright-server