Cursor can now control your computer. Agents can drive desktop apps like the iOS Simulator to test the features they build.

Computer Use lets agents click and type across any app on your computer, not just the browser. Cursor uses it on its own when a task needs it. You can also ask for it by @-mentioning Computer Use in your prompt. Update Cursor to try it:

oh that’s a good add. been waiting

how does it compare to computer use from codex and others?

It's about time! Thanks Cursor team!

@bot lets look at setting this up on the GrokBot Mini

Cursor we are in 2026 come on bro

A bit late guys, codex is too good at this.

Don't do this to me, I'm trying to have friends

“it should work now” go ahead, click it yourself

The real leap isn’t AI writing code faster it’s AI testing what it builds in the actual environment. That feedback loop could turn software development from a linear process into continuous iteration.

@martin_casado Hey guys. When did you last open Cursor?

@bot since in the background you run Cursor, can I ask you to use my computer too?

Reproducing bugs, verifying fixes, and catching regressions without human help matters more than just clicking through an app

My agent just opened the iOS Simulator, tested its own feature, and approved its own PR. I'm now the least productive person on my team and I'm the only person on my team.

Been using this with @bot for Xcode compilation

Agent builds the feature, agent tests the feature, agent says the feature works, now somewhere a QA engineer is laughing. Now the real question is: does it catch the bug it wrote, or just screenshots the happy path? haha

can i lock it to the simulator window or does it get the whole desktop

Can cursor also update Grok 4.7 It's the worst model I used in last 6 month. And it uses all my usage without delivering much. Thanks

cool until it opens my bank tab to test something

LMAO, how long do this take? I’m not trying to give you guys a hard time, but this took way too long

hiking app, main branch, one sentence. not that hard anymore

Desktop control is where agent memory starts to matter. Once it's clicking through your simulator and apps, it should remember what it learned about your setup next session, not rediscover it. I keep that in local files any agent can read, so Cursor and Claude Code share it.

Wasn’t this feature deployed on yesterday’s version?

this is huge for game builds. the agent can finally play-test what it just built instead of me doing it every round

@grok is it only for mac os?

The real win is the boring stuff. Agent builds the settings screen, then rotates the simulator, cranks Dynamic Type to max, and finds the button that falls off the edge. That's the bug humans skip.

and like Elon say : it's digital optimus

Why no native iOS sim? I added mine flawlessly into my app, makes iOS development so much easier

testing in the simulator is the part i'm really excited about! curious how often the agent decides on its own to go check the ui vs only when told to? this habit seems harder to train than the clicking part

@poteto Good

Big step for testing native apps. For web flows, many people still prefer to keep the agent in its own browser instead of the whole desktop. That’s what Hronaut does: a visible Chromium for Cursor, Claude Code and Codex, where you can pause the agent and take over anytime.

Would love the same for Grok Bot's computer. I'd rather the agent drive a machine in the cloud, safer and out of the way of mine, with mine as an option when I want it.

No.

finally