Ask HN: How to AI analyze what's on users' screen (with their consent)

I'm building an AI sales coach with automated time tracking.

What's a the best API/library to take screenshot every X seconds, send them to my favorite LLM to analyze it further.

What stack/pipeline are other tools like the below using?

https://highlightai.com/ https://www.focusbuddy.ai/ https://pointone.com/