Claude Code on iPhone: what actually works

By Techguyver28 September 2026 · 9 min read

Search "claude code on iphone" and you get a pile of guides for third-party tools, most written before Anthropic shipped its own answer.

Short answer: You cannot install the Claude Code command-line tool on an iPhone, because iOS does not allow a general-purpose terminal process to run in the background the way macOS, Windows, or Linux do. What you can do is reach a Claude Code session that is running somewhere else - in Anthropic's cloud, on your own computer through the official Claude app, or on your computer through an SSH or remote-control tool - and every real "Claude Code on iPhone" setup is one of those three.

That is worth saying plainly, because most of what ranks for this query is a guide for a third-party bridge tool, dated 2025 or early 2026, written before Anthropic shipped an official phone story. Some of it is now redundant. Some of it still solves a real gap. Here is what is actually true as of September 2026, official source by official source, plus the honest tradeoffs between them.

Option 1: the official Claude app, with cloud sessions

Anthropic's own Claude app for iOS has a Code tab. According to Anthropic's own documentation, tapping it lets you start a task against a GitHub repository that runs on cloud infrastructure, Anthropic-managed by default, so the session keeps working after you lock your phone and your laptop can be off the whole time.

Install the Claude app from the App Store and sign in with your claude.ai account.
Tap Code in the app navigation, or open claude.ai/code/new in Safari on your phone.
Pick a repository and branch, describe the task, and submit it.
Open the session later to check progress, answer a question Claude asks, or steer it - the same session is also visible from your laptop or any browser.

This is the closest thing to "Claude Code that lives entirely on my phone." The limit is the one you would expect: it needs your repository on GitHub and Anthropic's cloud environment set up first, per their cloud quickstart, and it is not your actual laptop, so a local file outside git, a running dev server, or a tool only installed on your machine is not reachable this way.

Option 2: Remote Control, for a session already running on your Mac

If the work needs your actual filesystem, your MCP servers, or a dev environment that only exists on your laptop, Anthropic's answer is Remote Control, which launched in February 2026. You start Claude Code on your computer as usual, and your phone becomes a window into that exact session.

On your Mac, run claude remote-control in your project directory, or type /remote-control inside a session that is already open.
Accept the one-time confirmation prompt.
Scan the QR code the terminal shows with your iPhone camera, or open the Claude app and find the session under Code - it shows a computer icon with a green dot when it is online.
Send messages from your phone. They land in the same conversation your terminal is running, and replies sync both ways.

Per Anthropic's own requirements page, this needs a Pro, Max, Team, or Enterprise plan (API keys do not work), and your laptop has to stay on with the claude process running the entire time - close the lid and the session goes offline until you bring it back. Anthropic is also direct that while your code execution stays local, "the session transcript, including your messages, Claude's responses, and tool activity, is stored on Anthropic servers" so it can sync across your devices.

Option 3: Dispatch, for messaging a task instead of driving a session

There is a third official route, and it answers a different question: not "let me see my session," but "let me hand off a task without opening a session at all." Dispatch is a persistent conversation in the Cowork tab of Claude Desktop. You message it from your phone, and Anthropic's docs say it decides whether the task is development work - in which case it spawns a Claude Code session on your Mac and you get a push when it finishes or needs approval.

Dispatch requires a Pro or Max plan and, same as Remote Control, needs your computer on with Claude Desktop running to actually do anything. Worth knowing before you lean on it for real work: Anthropic's own Computer Use documentation, which Dispatch sessions can use, calls it a feature that "can make mistakes" and recommends starting "with the apps you trust and not work with sensitive data" - reasonable advice for letting an agent click around your machine unattended, and a different risk profile than a session you are watching yourself.

Option 4: SSH from a terminal app

Before Anthropic shipped any of the above, this was the only way, and plenty of people still prefer it. Install an SSH client built for iOS - Blink Shell or Termius are the two most commonly used - and connect straight into a terminal session on your Mac, where the real claude CLI runs exactly as it does at your desk.

Start Claude Code inside tmux or screen before you disconnect, so the session survives you closing the SSH connection and reconnects to the same running process when you come back.
Mosh, rather than plain SSH, keeps the connection alive across a network change - walking from Wi-Fi to cellular does not kill the session the way SSH alone often does.
You are typing on a phone keyboard into a real terminal, which is the actual tradeoff: full CLI power, cramped input.

This is the most capable option in one specific sense: it is not a client for Claude Code, it is Claude Code, unmodified, with every flag and every local tool available exactly as installed. It is also the option with the least polish - no push notification when the agent stops and needs you, no lock-screen reply, nothing waiting for you except a terminal you have to remember to check.

Option 5: MightyMouse, for the moments you need to see the Mac, not just the agent

All four options above assume the thing you need from your phone is the Claude Code conversation itself. Often that is true. Sometimes it is not - the agent hit a dialog box nobody expected, a browser window it opened needs a look, or you just want to glance at the actual desktop rather than read a summary of it. MightyMouse turns your iPhone into a live screen, trackpad, keyboard, and microphone for your Mac, so instead of a chat client for one agent, you get the whole machine, with the agent's terminal as one of the things on it.

The part that solves the same problem Remote Control and Dispatch solve - a push the moment the agent needs you - works over MCP, the same protocol Claude Code already speaks to reach any tool. When a session calls a small set of relay verbs, the request reaches an app running on your Mac, which sends your phone a push. Tap it and, for a plain yes-or-no or a short pick-one, you can answer right from the lock screen without unlocking the phone. Behind that: the push notification itself carries no content, only an id - your phone fetches the actual question, and posts your answer, by talking to your own Mac directly, not through a server holding the text of your prompt.

Open Settings on your Mac and scan the code with your iPhone.
Safari opens a control page: drag to move the real cursor, tap to click, hold the mic button to speak.
When your coding agent calls the relay tool to ask something, your phone gets a push. Reply from the lock screen for a simple approve or a short pick, or open the control page to look at the actual screen and type a longer answer.

This works with Claude Code specifically, and with any other MCP-speaking agent the same way - Codex, Cursor, and VS Code among them - which is the honest reason to pick it over an Anthropic-only tool: it does not change shape when you switch agents. See how the relay mechanism works in more detail and exactly what does and does not travel through the push.

How the five compare

Official Claude app, cloud sessions. Best when the repo is on GitHub and you want work to continue with your laptop off. Cannot see or touch anything outside that repository.
Remote Control. Best when you already started a session at your desk and want to keep talking to it from the couch or on a walk. Needs Pro/Max/Team/Enterprise and your Mac to stay awake.
Dispatch. Best for handing off a task with zero setup at the moment you think of it. Pro or Max only, and it decides for you whether something counts as a coding task.
SSH with tmux/mosh. Best when you want the actual CLI, no substitute, and you do not mind a small keyboard. No push notification built in - you have to remember to check.
MightyMouse. Best when you need to see or touch the real Mac, not just read the agent's side of the conversation - and when your setup includes more than one AI agent, so a single Anthropic-specific app would not cover it.

The honest case against MightyMouse here: if all you ever want is the Claude Code conversation and nothing else on the Mac, the official app is simpler and needs nothing installed on the Mac beyond Claude Code itself. Reach for MightyMouse when the agent is only part of what you need from the machine, not the whole job.

Why the CLI itself cannot run on an iPhone

This is the part worth being direct about, since it is the actual question behind the search. iOS does not let a regular app keep an arbitrary background process alive and reachable over a network the way a Mac, a Windows PC, or a Linux server does - that model is the whole reason SSH, Remote Control, cloud sessions, and Dispatch all exist as ways to reach a machine that can run the CLI, rather than as ways to run it on the phone directly. None of Anthropic's own documentation describes an iOS build of the Claude Code binary, and nothing above works around that; each one reaches a session running somewhere that isn't the phone.

See your Mac and answer your agent from your iPhone, not just the chat.
Your iPhone as a screen, trackpad and mic for your Mac. No card to try it.

Common questions

Can I install Claude Code on my iPhone?

No. The Claude Code CLI needs a background process an iOS app cannot keep alive, so it only runs on macOS, Windows, or Linux. Every phone option reaches a session running on one of those.

What is the difference between the Claude app's Code tab and Remote Control?

The Code tab can start a brand new cloud session that runs on Anthropic's servers with your laptop off. Remote Control instead connects your phone to a session already running on your own computer, so your files and local tools stay available and your computer has to stay on.

Do I need a paid Claude plan to use Claude Code from my phone?

For Anthropic's own official routes, yes - Remote Control needs Pro, Max, Team, or Enterprise, and Dispatch needs Pro or Max. SSH into your own computer needs no Claude subscription tier beyond whatever your Claude Code setup already uses, and MightyMouse's phone control needs no AI subscription at all, only your own AI account or its built in option.

Does Claude Code send my code to a server when I use it from my phone?

It depends on the route. Cloud sessions run on Anthropic-managed infrastructure by design. Remote Control keeps execution and filesystem access on your own machine, though Anthropic stores the session transcript to sync it across devices. An SSH connection or MightyMouse's relay both keep your code on your own machine and only pass the conversation between your two devices.

Can I use MightyMouse and the official Claude app together?

Yes. They answer different questions and do not conflict - use the Claude app or Remote Control for the agent conversation itself, and MightyMouse for the moments you want to see the actual Mac screen or drive it directly.

If your agent needs more than a chat window: MightyMouse puts your Mac on your iPhone as a live screen, trackpad and mic, with free local voice transcription and a lock-screen push the moment it needs you.