Do you use coding agents from the terminal or from the app?

I’m curious how people are actually using coding agents today.

Do you mostly use them directly in the terminal, or through an app/interface like Cursor, Claude Code, Codex, etc.?

What’s your preferred setup?

101 views

Add a comment

Replies

Best

i personally prefers the terminal version

An IDE, preferably Cursor or VS code

 Do you still take a look at your code, especially in vscode ?

 yes I do, I use AI for debugging and Understanding issues, on my code and direct it to follow the algorithmic flow

Terminal for anything touching more than about three files, IDE for anything I already suspect is wrong. That split isn't taste, it's about where the review happens. In the terminal the review surface is the diff, so I have to open git and actually read it. In the IDE the review happens as it types, and I've caught myself nodding along at changes I never really read. The RAM cost is real but it's the smaller one. The bigger one is that a terminal agent is much easier to leave running unsupervised, which is exactly how we ended up with adapters whose tests only checked our own output shape.

An agent assumes autonomous work? Without you being there 100% of the time? I don't trust it enough to do that. I work through Terminal because usually I have Android Studio, VS Code, and Xcode open (building a KMP app), and I need something that can oversee all of them.

the terminal hurts my eyes and the claude code app is pretty nice these days