It looks when you ask
There is no background monitoring, no always-on recording, and no timer taking screenshots. A capture happens at the moment you hold the hotkey and ask a question - and at no other time.
Screen-aware AI for Mac
Ask About What You Are Looking At
Hold a key, ask out loud, and Crodo answers about whatever is on your screen - the error, the dashboard, the document, the design. No screenshotting, no uploading, no switching tabs. And it only ever looks at the moment you ask.
Free · Screenshots never stored · macOS 14.2 or later · Apple silicon
Screen question
Asked once, not watched
You said
What is this error actually telling me, and what should I try first?
Captured: the display under your cursor · Crodo's own windows excluded
Crodo
What it means
The request never reached the server - the port is closed, not the code failing.
Try first
Check the dev server is running, then confirm the port in your config matches.
How it behaves
It does not watch your screen.
“AI that sees your screen” usually sounds like something running in the background, always looking. That is not what this is. Crodo takes a single capture at the instant you ask a question, and does nothing in between.
The screen you are actually on
With more than one display, Crodo captures the one under your mouse cursor rather than guessing. Your cursor is in the shot too, so “what is this?” works when you are pointing at something.
Crodo stays out of the picture
The app's own windows are filtered out of the capture, so the assistant never ends up looking at itself instead of your work.
Nothing is kept
The screenshot is held in memory long enough to answer your question and is never written to disk or saved to your account. Only the text of the question and answer becomes conversation history.
How it works
One key, one question.
The whole interaction is shorter than taking a screenshot would be.
01
Hold the screen-aware hotkey and ask your question out loud - “what does this error mean?”, “summarise this page”, “what is wrong with this layout?”
02
Crodo captures the display under your cursor, excluding its own windows, and downscales the image before it leaves your Mac.
03
Your spoken question is transcribed and sent with the screenshot to a vision model.
04
The answer streams back into Crodo, ready to read or copy into whatever you were doing.
Where it earns its place
Best when copying is the hard part.
Anything you can see but cannot easily select is where this stops being a novelty and starts saving real minutes.
Errors and stack traces
Point at a red wall of text and ask what it means and what to try next - no copying, no pasting, no losing your place in the terminal.
Dashboards and charts
Ask what a graph is showing, which number moved, or what looks off in an analytics view that has no export button.
Dense documents and PDFs
Summarise a contract clause, a spec, or a research paper that is visible on screen but painful to select and copy.
Design and UI review
Ask what is unbalanced in a layout, whether spacing looks consistent, or how a screen reads to someone seeing it for the first time.
Unfamiliar software
Open a tool you have never used, ask where the setting you need lives, and get an answer about the actual window in front of you.
Anything you cannot select
Images, video frames, canvas apps, remote desktops, screenshots someone sent you - if you can see it, you can ask about it.
What happens to the image
Used to answer, then gone.
A screenshot of your work is about as sensitive as data gets, so the specifics matter more than a reassuring sentence. Here is exactly what happens to it.
Never written down
The image is held in memory only for as long as it takes to answer. It is not written to disk on our server, not saved to your account, and not attached to your history. What is kept is the text of your question and the answer - which you can delete.
Where it does go
Being straight about the trade: the screenshot is downscaled and sent to a cloud vision model to be read. It has to leave your Mac for this to work at all, and screen questions need an internet connection. If nothing may ever leave your machine, this is not the feature for you.
You control when
Capture happens on your keypress and nowhere else, and the whole feature sits behind macOS Screen Recording permission, which you can revoke in System Settings at any time. Meeting recording does not use this permission - only screen questions.
Questions
The things people ask first.
Does Crodo watch my screen all the time?
No. Crodo captures your screen only at the moment you hold the hotkey and ask a question. There is no background monitoring, no continuous recording, and no periodic screenshotting. Between questions, nothing is being looked at.
Are my screenshots stored?
No. The screenshot is held in memory only for as long as it takes to answer, and is never written to disk on our server or saved to your account. Only the text of your question and the answer is kept, as conversation history you can delete. The image is sent to the AI model provider to be processed and is not retained by Crodo afterwards.
Which screen does it capture if I have two monitors?
The one your mouse cursor is on. That is almost always the display you are working on, so you do not have to think about it. Your cursor appears in the capture, which is what makes questions like "what is this thing I'm pointing at?" work.
Will Crodo's own window show up in the screenshot?
No. The app filters its own windows - the recording pill and the main window - out of the capture, so the assistant never blocks or reads itself instead of your work.
What permission does this need?
Screen Recording permission, granted in System Settings → Privacy & Security. macOS asks the first time you use the feature. Crodo uses ScreenCaptureKit, Apple's current capture API, which is why it appears properly in that list. Note that meeting recording does not need this permission - only screen questions do.
Does it work offline?
No. The screenshot and your question are sent to a cloud vision model, so screen-aware answers need an internet connection.
Can it read text in images and screenshots?
Yes. Because the model sees the pixels rather than the underlying text, it works on anything visible - screenshots someone sent you, text inside images, video frames, canvas-based apps, and remote desktop sessions where copying text is not possible.
Is this different from taking a screenshot and uploading it to ChatGPT?
Same idea, far less friction. No screenshot key, no finding the file, no switching to a browser tab, no upload. You hold a key, ask out loud, and the answer arrives where you are working - and because it is the same assistant, it also has your memory, tasks, and connected apps in context.
Keep going