Cursor on Phone: What Mobile Actually Does

Yes, you can use Cursor on your phone, but it is a web-based agent dashboard, not the desktop IDE and not a terminal. As of 2026-06-02, Cursor has no native iPhone or Android app. You open cursor.com/agents in your mobile browser, install it as a Progressive Web App, and from there you start cloud agents, approve their steps, and review the pull requests they open. The actual code editing still happens back on a desktop. If what you actually want is an interactive coding environment on your phone, that is a different shape of tool, covered at the end.

This is the Cursor-specific companion to our broader AI coding agents on mobile pillar, and a sibling to the head-to-head in Cosyra vs Cursor. Cursor and a mobile terminal solve different halves of the same problem, so this guide sticks to one question: what does "Cursor on your phone" really get you, and where does it stop? If it is plain VS Code on your phone you are after rather than Cursor's agent layer — Cursor is a fork of VS Code — that guide covers what the browser build (vscode.dev) can and cannot do. Windsurf (renamed Devin Desktop in 2026) is another AI fork of VS Code in the same desktop-only spot; running Windsurf's agents from a phone follows the same cloud-container path this guide lays out.

What "Cursor on phone" actually is

Cursor announced agents on web and mobile in a June 2025 blog post, framing it as agents that "can write code, answer complex questions, and scaffold out your work" from a browser. The mobile half is delivered as a PWA, not a native app. Cursor's own documentation says to install the web interface to your home screen for "a native app experience on iOS or Android." The compute happens in a cloud VM; your phone is the remote control.

That distinction matters because most people typing "cursor on phone" into Google are picturing the desktop IDE shrunk onto a screen: the editor, the Tab autocomplete, the inline diff. That is not what you get. What you get is a place to kick off an autonomous agent and supervise it. You write a task, the agent runs in the background, and you come back to a diff and a pull request. Cursor's docs and blog are consistent on this: review diffs, review and create pull requests, add follow-up instructions, run multiple agents in parallel. The editing surface is still the desktop. Approving a diff that way is fine for a quick sign-off; reviewing a pull request more thoroughly from your phone — checking it out, running its tests, AI-reviewing the diff — needs a real shell rather than a review surface.

How to set up Cursor on your phone

There is nothing to install from an app store. The whole setup is opening the web interface and pinning it to your home screen so it behaves like an app:

1. Open the web interface. In Safari on iOS or Chrome on Android, go to cursor.com/agents and sign in with your Cursor account.
2. Install it as a PWA. On iOS, tap the share button, then Add to Home Screen. On Android, open the Chrome menu, then Install App. You now have a full-screen icon with no address bar.
3. Launch or open an agent. Type a task to start a cloud agent, or tap into a running one to approve a step, send a follow-up, and read the diff it produced.

The one thing worth setting up before you leave your desk: connect the repositories you want the agent to touch. Cloud agents work against a connected Git repo, clone it into their VM, and push a branch. If the repo is not connected, the phone is a dead end. You cannot fix that from Safari as easily as from the desktop.

What you can and can't do from your phone

Here is the honest split between what Cursor's mobile web surface handles well and what it leaves on the desktop. We pulled the "can" column from Cursor's own blog and docs, dated 2026-06-02:

Read that table top to bottom and the pattern is clear: the mobile experience is supervision, not authorship. That is a deliberate design choice, and for a lot of workflows it is the right one. Launch a bug fix from the coffee-shop queue, approve the agent's plan, and merge the PR from the couch that evening. You never touched a keyboard for the code itself.

The honest gap: agent control vs an interactive terminal

Here is an opinion the Cursor team would not share: agent-only control is the wrong default for phone coding. Cursor has bet that on a phone you should supervise an autonomous agent and nothing else. We think that bet is half right. Plenty of the time on mobile you do want the agent to drive, but the other half of the time you want to run one command. Pull the latest, restart a dev server, check why a test failed, kill a stuck process, paste a one-off curl. Cursor's mobile surface has no terminal, so all of that waits until you are back at a desk.

We build Cosyra for that other half. It is a native iOS and Android app wired to a persistent Ubuntu 24.04 container on Azure, with Claude Code, Codex CLI, OpenCode, and Gemini CLI already installed. You type into a real shell, not a prompt box. You can run an agent interactively, watch its output stream, ctrl-C it, and run a build right after, the same things you would do in a desktop terminal, on the phone in your hand. The container persists across app restarts and device switches, so a session you start on the train is still there on the couch.

To be precise about what this is and is not: Cosyra does not run the Cursor IDE, and we would not pretend it replaces Cursor on the desktop. They aren't direct substitutes. Cursor owns the desktop IDE workflow, a mobile terminal owns the type-a-command-from-anywhere workflow. The realistic setup for a lot of developers is both: Cursor's web agent to supervise long-running tasks, and a mobile terminal for hands-on work when you are away from the laptop. We unpack the full trade-off in Cosyra vs Cursor.

Who should pick what

Match the tool to how you actually work on mobile:

• Use Cursor's web agent if your mobile coding is mostly supervision: you queue tasks for autonomous agents, approve their steps between meetings, and review PRs. You already live in Cursor on the desktop and want the same agents reachable from a browser. This is the strongest fit for Cursor's mobile surface, and it is genuinely good at it.
• Use a mobile terminal (like Cosyra) if you want to type commands from your phone: run a build, restart a server, drive a CLI agent interactively, or tail a log, on a persistent Linux box reachable from a native app.
• Use both if you do long-running autonomous work and hands-on work from the same phone. They overlap less than you would think: one is a supervisor's dashboard, the other is a shell.

If you are weighing the broader field of agents that run from a phone (Claude Code, Codex CLI, Gemini CLI, OpenCode), start with our AI coding agents on mobile pillar, or the hands-on Claude Code on your phone walkthrough for the interactive-terminal path end to end.