Cursor's mobile app for coding agents: fine for a nudge, useless for real review
Cursor's mobile app is a solid remote control for starting and nudging a background coding agent — but reviewing real diffs on a phone falls apart.
Short version: the Cursor mobile app is genuinely useful for kicking off and babysitting a background agent when you're away from your desk, but the moment the agent needs a real decision, you're stuck squinting at a diff on a 6-inch screen. I ran one from my phone on the train for about 40 minutes. It works. It also made me appreciate my monitor.
Cursor, from Anysphere, shipped a mobile companion that lets you monitor and steer your coding agents remotely instead of being chained to the desktop IDE. TechCrunch covered the launch in the original report, and the dev-community reaction has been loud because async background agents keep eating more of the workflow.
Does firing an agent from your phone actually work?
Yes, for the narrow case of starting a task and checking progress. I typed a prompt — "add pagination to the /events list endpoint and update the tests" — hit go, and watched the agent churn while I sat in a tunnel with two bars of signal. The status updates streamed fine. When the agent asked a clarifying question, I could answer in one line and it kept going.
That's the sweet spot: a task you can describe in a sentence, that doesn't need you to read much back.
What broke?
Review broke. The agent produced a diff touching four files, and reviewing that on a phone is miserable. You can't scan a 200-line diff on mobile the way you scan it on a wide monitor. I ended up approving on faith, which is exactly the habit these tools should not encourage.
The second problem is context switching. On desktop I can flip to the failing test, check the actual error, tweak the prompt. On mobile every one of those steps is three taps and a scroll. The friction quietly pushes you toward rubber-stamping whatever the agent did.
Also worth noting: flaky signal means the session state gets weird. Twice the app showed a stale "running" state after my connection dropped, and I had to reopen to see the agent had actually finished.
So what's it actually good for?
The honest use case is triage, not authorship. Kick off a long-running task before you leave, glance at it from the couch, answer a quick clarifying question, and do the real review when you're back at a keyboard. Treating the phone as a full agent-driving surface is where the workflow falls apart.
One implication nobody's saying out loud: mobile agent apps only make sense if you already trust the agent enough to not read its output carefully. That's a big assumption for anything touching production code.
Should you use it?
Install it if you already lean on Cursor's background agents and want to start or unblock a task while away from your desk. Skip it if you expected to actually review and merge from your phone — that part isn't there, and honestly the screen size means it may never be. It's a remote control, not a workstation. I'll keep it for the train-station nudge and nothing more.