Cursor has launched a native iOS app in public beta, giving developers a way to work with AI coding agents without being tied to a laptop. The app lets you start agents, track their progress, review diffs, and merge pull requests directly from your phone.
The release marks a real shift in how developers can interact with AI tools. Until now, using Cursor meant sitting at a computer. The mobile app changes that by connecting your phone to agents running either on your local machine or in Cursor’s cloud infrastructure, so the work keeps moving even when you step away from your desk.
The app is available now on all paid plans. Through July 5, 2026, users get 75% off Composer 2.5 runs made through the mobile app.
The core idea is simple: moments of inspiration don’t always happen at a desk. You open the app, pick a repo, describe what you want, and launch an agent. You can use voice input to describe ideas out loud, apply slash commands to guide the agent, and choose from any of Cursor’s supported frontier models. If an agent is already running on your computer, the Remote Control feature lets you keep directing it from your phone. There’s also a setting to keep your machine awake and reachable while you’re away.
The Cursor team says they’ve been using the app internally for a range of tasks, and a few workflows stand out as particularly useful:
- On-call incidents: When you get paged during lunch, you can launch an agent to investigate and propose a fix. By the time you’re back at your desk, a pull request may already be waiting.
- Customer bug reports: If a time-sensitive issue comes in while you’re away, you can start an agent from your phone to reproduce the problem and start working toward a fix.
- Design feedback from other apps: When you spot user feedback on X or another platform, you can screenshot it, annotate it, and send it to an agent as visual context to kick off design or UI changes.
Once an agent is running, you don’t need to stay in the app. Cursor sends push notifications and shows Live Activities on your lock screen to keep you updated when an agent finishes, needs input, or has something ready to review. Cloud agents also produce demos, screenshots, and logs, so you can check their work without switching to a computer.
The cloud agent setup is worth understanding. These agents run in isolated virtual machines with full development environments. Because they operate asynchronously with their own tools and resources, they can run longer and iterate further than local agents without needing you to intervene. You can send a local plan to a cloud agent, or shift an active agent to the cloud mid-task. When the work is done, you can pull the session back to your local machine to test changes before merging.
This local-to-cloud handoff is one of the more interesting parts of the release. Most AI coding tools force you to choose one environment. Cursor’s approach lets you move work between your machine and the cloud depending on what the task needs, and the mobile app sits in the middle of that workflow as a control layer.
Looking ahead, the team says the gap between running agents locally and in the cloud will shrink over time. They’re also working on repo-less chats, which would let you kick off tasks that don’t need codebase context. Some teams are already using Cursor with MCP integrations to query Datadog logs, summarize Slack activity, and handle tasks that go well beyond writing code.
The broader trend here is that AI coding assistants are moving from being desktop tools to something closer to always-on infrastructure. Cursor for iOS is an early sign of what that looks like in practice: agents that run in the background, report back when ready, and don’t need a developer staring at a screen the whole time.
