Cursor ships an iOS app, moving the coding agent off the desktop
Cursor released a public beta of its iOS app on June 29, 2026, letting developers launch cloud agents, drive their desktop sessions remotely, review diffs, and merge pull requests from a phone. It is available on every paid plan.
Cursor put its coding agents on the phone on June 29, 2026, releasing a public beta of a native iOS app. The pitch is not mobile editing, which few developers want, but mobile supervision: starting cloud agents, steering the ones already running on a desktop, and approving their work from anywhere. It reads as the logical next step for a tool that has spent the past year repositioning around agents rather than the editor.
What shipped
Cursor for iOS is available on every paid plan and, for now, on iOS only, with no Android or web client announced. The app does three things. It launches cloud agents directly, so a task can start without a machine open. It connects to desktop agents through a Remote Control feature, so a session already running on a workstation can be driven from the phone. And it handles the review loop: developers can read diffs and generated artifacts (demos, screenshots, logs), annotate screenshots, and merge pull requests from the app.
The interaction details lean into the form factor. Tasks can be dictated by voice, guided with slash commands, and tracked through push notifications and lock screen Live Activities while an agent works. Handoff runs both ways between local and cloud agents, so a job can move between the desk and the phone without restarting. Cursor framed the launch around moments away from the keyboard: on-call incident response, fixing a customer bug, or acting on feedback while out, with the company writing that until now developers have “worked around the limits of their local machines, keeping laptops half-open and caffeinated everywhere they go.” A promotion runs alongside the beta, with 75% off Composer 2.5 runs in the mobile app through July 5, 2026.
For a working developer, the practical change is the dead time. The agent model already decoupled writing code from supervising it, but supervision still assumed a laptop. An app that can kick off a run, surface the diff, and merge it turns the on-call hour and the commute into time an agent can be pointed at, which is the actual value here rather than typing code on a touchscreen.
Where this lands in the market
Cursor is not first to this idea, and that is the point. Anthropic’s Claude Code and OpenAI’s Codex both already offer ways to reach agents from a phone, and the head of Claude Code has said most of his coding now happens on mobile. The mobile client is becoming table stakes for any tool that sells autonomous agents, because an agent that runs for minutes without input is one a developer should be able to check on without being chained to a desk.
The launch also fits Cursor’s own trajectory. Since the agent-first turn it took with Cursor 2.0 in late 2025, the product has steadily moved its center of gravity from the file you are editing to the work the agent is doing, and the mobile app extends that reach to wherever the developer happens to be. It also arrives at a moment of unusual scrutiny for the company, weeks after the reported SpaceX acquisition put Cursor under new ownership, which raises the stakes on shipping cadence rather than lowering them.
What’s worth watching
- Whether mobile changes who supervises agents. If approving a merge from a phone becomes routine, the reviewer and the author can drift further apart in time and place. The teams that benefit will be the ones whose review discipline survives the convenience.
- The beta-to-stable gap. Public beta on iOS only, with no Android or web, leaves real coverage questions. How fast the platform list grows, and whether usage limits on mobile differ from the desktop, will decide how much of the workflow actually moves.
- Cost behavior on the phone. Background and cloud agents bill for compute whether or not anyone is watching the diff. A discounted-run promotion is a nudge to start more runs; the durable question is what mobile-initiated agent usage does to a team’s bill once the discount ends.
The quieter shift is that the phone is becoming a control surface for software work rather than a place to read about it. Cursor is betting that the next unit of developer productivity is not the keystroke but the approval, and that the approval should travel. Stackmaven’s follow-up coverage will revisit adoption and platform coverage on or around September 28.