Cursor has released a public beta of its iOS app, available to all paid subscribers, adding mobile access to the AI-powered coding environment alongside the ability to launch and manage always-on agents from a phone or tablet. The release was noted in the Cursor changelog on 28 June 2026.

The core addition beyond a mobile interface is agent orchestration on the go. Cursor's always-on agents — long-running, asynchronous processes that can execute tasks, run tests, and iterate on code without requiring the developer to be at a keyboard — can now be kicked off and monitored from iOS. That changes the practical footprint of the feature considerably. An agent doing a multi-hour refactor or a background sweep of a large codebase no longer demands desktop presence to initiate or check.

For developers already using Cursor's agentic workflows, the friction point has generally been the handoff between where you start a task and where you follow its progress. Mobile removes that seam. You can queue work from wherever you are, watch the agent's output stream, and intervene or redirect without needing to be at a workstation. It is a narrowly scoped capability, but it fits naturally into how longer-running AI tasks actually get used — in bursts of attention spread across a working day, not in one unbroken sitting.

The public beta is gated to paid tiers, which follows a pattern Cursor has used for earlier feature rollouts: limit initial exposure to the subscriber base most likely to generate structured feedback, then broaden. That also means the feature is functionally available now to most active Cursor users, not a closed waitlist.

Worth flagging: the beta label matters here. iOS development environments carry their own constraints — sandboxing, background execution limits, connectivity assumptions — that are distinct from what Cursor manages on desktop. How the app handles an agent that runs long enough to exceed iOS background refresh windows, or how it behaves on a flaky mobile connection mid-task, are the kinds of edge cases that will determine whether this becomes a daily-use tool or a convenience for lighter interactions. Beta users on paid plans are effectively the stress test.

The broader trajectory is legible enough. As AI coding assistants move from autocomplete toward genuine multi-step autonomy, the interface layer starts to matter less than the ability to delegate and then check back in. A desktop IDE is the right place to author and review code at depth. A mobile client for agent management is a different surface entirely — closer to a job queue dashboard than a coding environment. Whether Cursor positions the iOS app primarily as the latter, or tries to expand it toward something closer to a full mobile IDE, will be worth watching as the beta progresses.