No switching to a browser, no pasting screenshots into a chatbot, no losing your place. Guidelet reads whatever is on your screen, answers in two sentences, and draws a box around the exact control you need — on top of the app itself.
No plugins. No per-app setup. Nothing to integrate. Guidelet reads pixels, not APIs — so it works in anything you can see, including the app nobody has written a tutorial for.
Recorded in Microsoft Word and CapCut · no edits
Four context switches to press one button.
You never looked away.

One shortcut, over whatever is in front — the bar names the app it is reading before you type a word. Word is still there, untouched, behind it.

Not a description of where the button is. A box on the button, numbered in the order you touch it.
Word keeps its window and stays usable. The answer takes the side, the box stays on the button, and you carry on in the document you were already in.
Nothing here was built for a video editor. The same thing that finds a button in Word finds one in CapCut, Blender, or software with twelve users — because it looks at your screen the way you do.

Questions stay filed under the app you asked them in, so Word answers never mix with CapCut ones.
Keep an answer and its boxes for later. Reopen it the next time you need the same steps.
Drag a box around one panel and ask about that alone, instead of the whole screen.
It searches on its own when the app is niche, and lists the sources it used under the answer.
Send an answer with its highlighted screenshot to someone else, or save it to disk.
Ask the next question without leaving the app or losing the boxes already on screen.