Cursor Mobile: iPhone, iPad, Android Status & How to Use It

The app is for directing agents away from your desk: starting runs, checking progress, reviewing pull requests and merging. The table lists the platforms, plans and limits as Cursor documents them.

Cursor's release notes and announcements say what changed. They don't say how a team should act on it. The steps below are our suggested order, and the next section lists what to watch out for along the way.

1. 1Use mobile for supervision and follow-up. Don't approve merges blind from a phone.
2. 2Keep high-risk edits behind desktop review, tests and code-owner approval.
3. 3Write down which actions are safe from mobile: comment, redirect, inspect status and request evidence. Each can be undone in about a minute, unlike a merge.
4. 4Revisit those rules when Android ships or Cursor changes which plans include the app.

Each item below is either a fact about the release, such as its beta status or plan access, or our advice for using it safely. Plan around each one, and recheck the facts when Cursor ships an update.

• Mobile doesn't replace a full diff review. A run you start on your phone waits for you on the desktop, where you can read the diff at full size.
• Cursor has not shipped an Android app. Wait for an official source before you plan around one.
• Apply the same policy to secrets, production access and approvals as you do for desktop and cloud agents.

There is no separate invite or mobile entitlement: anyone who can sign in and reach cursor.com/agents can use the app, within the plan limits above. The Cursor for iPadCursor's larger-screen mobile app for starting and directing cloud agents, reviewing full pull requests and marking up changes; it is not a full local IDE. Press Enter for the full definition. guide covers the iPad's larger review layout and Apple Pencil workflow.

Run the Agents WindowCursor's surface listing your agent runs; open a run to read its diff and search transcripts from the command palette. Press Enter for the full definition. next to your native IDE as a sidecar. In Cursor's mobile workshop, the presenter's premise was that Cursor already writes Swift and Kotlin well and lacks the native toolchain's work: building, running the simulator, reading logs and debugging. The sidecar pattern bridges that gap, with information moving in both directions.

The Agents WindowCursor's surface listing your agent runs; open a run to read its diff and search transcripts from the command palette. Press Enter for the full definition. is Cursor's agent-first desktop surface, which Cursor built from scratch for Cursor 3An older release label for Cursor's multi-agent interface; current product guidance calls the active surface the Agents Window. Press Enter for the full definition. instead of on its VS Code fork. For mobile work, keep it open beside Xcode or Android Studio and switch between them. To open it from the editor, press ⌘+⇧+P and run Open Agents Window.

On iOS, the integration path is the MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. server built into Xcode 26.3 and later. Apple ships a binary, xcrun mcpbridge, that translates MCP messages into Xcode's internal XPC layer, and Cursor adds it like any other MCP server. Xcode has to be running with your project open. Its 20 tools cover what Cursor can't do on its own for an iOS project, such as building, running tests, reading build logs and rendering SwiftUI previews.

Android: ACP, MCP plugins or both

The workshop presenter showed three ways to connect Cursor to Android Studio, and you can mix them. The presenter still prefers the sidecar, meaning the CLI or the Agents WindowCursor's surface listing your agent runs; open a run to read its diff and search transcripts from the command palette. Press Enter for the full definition., over chatting inside the IDE, but the in-IDE options are there when you want them.

For the MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. route, the presenter enabled the 'MCP server' pluginA Cursor marketplace package that bundles MCP servers and skills (sometimes sub-agents and hooks); one click installs all of it into your Cursor instance. Press Enter for the full definition. for broad IDE functionality, plus a separate 'debugger MCP server' plugin. The debugger plugin exposes fine-grained control: line breakpoints, conditional breakpoints, start, step out and step in.

React Native plus Expo: the lightest path

In the workshop, React Native was the smoothest of the three. Run the Expo server and it hot-reloads your changes, so you skip the build-and-run step that the Xcode and Android Studio MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. bridges exist to solve.

If a framework already gives you fast hot reload, skip the heavier MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. loop.

Pick by the job, because the two don't overlap. Cursor for iOS is the iPhone and iPad app for directing agents while you are away from your desk. The Xcode and Android Studio MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. Press Enter for the full definition. bridges are for building mobile apps at your desk, with the Agents WindowCursor's surface listing your agent runs; open a run to read its diff and search transcripts from the command palette. Press Enter for the full definition. open beside the native IDE.

If you ship an iOS app, start with the Xcode bridge. Its RenderPreview tool captures a SwiftUI preview as an image, so the agent can compare a UI change with what the preview shows. The app plays no part in that loop. Cursor's docs say it isn't an IDE: it starts, follows and redirects agent runs, and it reviews and merges their pull requests.

Decisions you can reverse are safe from a phone. Merges should wait for a desk, unless your team can roll one back quickly.

The steps above name the reversible ones: commenting, checking status, asking for evidence and redirecting a run. A merge can't be taken back in a minute, yet the app can merge a pull request as well as comment on it. That is why the line has to come from team policy.

Holding a merge until you reach a desk costs little. The run you were following on the phone is the same run in the desktop Agents WindowCursor's surface listing your agent runs; open a run to read its diff and search transcripts from the command palette. Press Enter for the full definition., with the full diff ready to read.

The July 29 release widened what that policy has to cover. The app now works with Bitbucket and Azure DevOps repositories as well as GitHub and GitLab. When one chat opens several pull requests, you can now open every one of them instead of only the last, so one session on a phone can reach more merges than it could in June.

Your rollback path decides which merges qualify for that exception. If your team can revert a merged change within minutes, a small merge from a phone carries little risk. If undoing it is slow or risky, leave that merge for the desk.