Claude Code on iPhone: 4 Real Ways to Run It Remotely

There is no standalone Claude Code on iPhone app that runs the agent on your phone. Claude Code is a terminal tool that needs a real machine. But you can absolutely drive it from an iPhone — over SSH into a Mac, Linux box, or VPS, or through Anthropic's official Remote Control feature.

Remote Control behaviour, session limits, and mobile-app boundaries verified 31 July 2026 against Anthropic's Claude Code documentation.

That distinction matters more than it sounds. Almost every question about running Claude Code on iPhone is really two questions stacked on top of each other: where does the agent run? and what am I looking at it through? The answer to the first is always a computer with a filesystem, a shell, and your repo checked out. The answer to the second can genuinely be your phone. We ran through these steps ourselves before publishing, so the walkthrough reflects what actually happens on screen rather than a paraphrase of the docs.

Once you separate those two things, the whole topic gets simple. Your iPhone is a window. Something else is the engine.

Why Claude Code on iPhone can't work the way you'd expect

The agent needs a general-purpose shell, arbitrary process execution, and a shared filesystem. iOS grants apps none of those by design, so no patch is coming. Every workable route puts the engine elsewhere and treats the phone as a window.

Claude Code is a command-line agent. It reads and writes files, runs your test suite, greps your codebase, executes git commands, and spawns processes. iOS does not give apps a general-purpose shell with arbitrary process execution and access to a shared filesystem. That's not an oversight Anthropic will patch — it's how the platform is designed.

So a native app that installs Claude Code onto your iPhone and lets it hack on a repo stored on the phone isn't a thing, and you should be suspicious of any post claiming otherwise. If you're new to the tool itself, our explainer on what Claude Code is covers the terminal-first model in more detail.

What is real: every practical route to Claude Code on iPhone works by putting the agent somewhere else and giving your phone a way to talk to it.

The routes that actually work

Four honest options, trading setup effort against reliability: SSH plus tmux into a machine you own, the same thing against a rented VPS, Anthropic's official Remote Control feature, and the chat-only Claude app that edits nothing.

There are four honest options, and they trade off very differently on setup effort and reliability.

1. SSH plus tmux — the most reliable route

This is the route that has worked for years and will keep working. You need a machine that stays on (a Mac at home, a Linux desktop, a Raspberry Pi, a VPS) with Claude Code installed and authenticated, plus an SSH client on your iPhone.

The one non-negotiable ingredient is a terminal multiplexer. Without tmux or screen, your Claude Code session dies the moment your phone sleeps, your train enters a tunnel, or iOS decides to suspend the app in the background. With it, the session keeps running on the server and you simply reattach.

# From your iOS terminal app
ssh user@your-machine
 
# First time: start a named session
tmux new -s claude
claude
 
# Phone slept, signal dropped, app got killed? Reconnect and reattach:
ssh user@your-machine
tmux attach -t claude

That's the whole trick. tmux new -s claude creates a persistent session named claude; tmux attach -t claude picks it back up exactly where you left off, mid-agent-run, with all the output still there. Detach deliberately with Ctrl-b then d.

A few things worth doing once, on the server side, before you rely on this:

1. Install and authenticate Claude Code on the machine while you're sitting at it — logging in via a browser flow from a phone terminal is miserable. Our guide to authenticating Claude Code in the terminal walks through it.
2. Set up SSH key auth rather than passwords. Typing a long password on a phone keyboard, repeatedly, will make you quit within a week.
3. Turn off sleep on the host machine, or your engine goes away exactly when you need it.
4. Pick a terminal app that supports Mosh or aggressive keep-alives if your connection is flaky.

2. A cloud VM or VPS

Same mechanics, different host. If you don't have a machine that's reliably awake and reachable, rent one. A small cloud instance with your repo cloned onto it is a perfectly good home for Claude Code, and it removes the "is my laptop lid closed?" failure mode entirely.

The tradeoff is that your code now lives on that server, you're paying for it monthly, and you have to keep it patched. For a side project you poke at from the couch, that's fine. For work code, check your employer's policy first.

3. Remote Control — the official answer

Anthropic ships a feature called Remote Control that is the closest thing to a real Claude Code on iPhone experience. You start a session on your machine with claude remote-control (or claude --remote-control, or /remote-control inside a session you're already in), and it shows you a session URL and a QR code. Scan the QR code with your phone and the session opens in the Claude mobile app; or open claude.ai/code in any browser.

The key detail — and the reason this doesn't contradict anything above — is that Claude still runs on your machine. Your filesystem, your MCP servers, your project config all stay local. The phone is a window into that session, not a second engine. We go deeper on the setup and its quirks in our piece on Claude Code Remote Control.

It's genuinely good. It reconnects automatically when your laptop wakes up, it can send you a push notification when a long task finishes or when Claude needs a decision, and you can send images from your phone straight into the session. The limits are honest ones: the local claude process has to keep running (close the terminal and the session ends), an outage longer than roughly 10 minutes ends the session, and some commands like /plugin and /resume are terminal-only. Server mode supports up to 32 concurrent sessions if you need more than one.

Availability and plan requirements move around — Remote Control launched as a research preview and the rollout has shifted more than once. Check the official Claude Code docs for the current state rather than trusting a blog post's snapshot, including this one.

4. The Claude mobile app is not Claude Code

This trips people up constantly. The Claude app on the App Store is the chat app. It's excellent — you can ask it to explain a stack trace, draft a function, or reason through a design. Currently it runs the same frontier models you'd use elsewhere, such as Claude Opus 5 for the heavy thinking.

But by itself it cannot edit your repo, run your tests, or execute a single command. It has no shell and no filesystem access. Our Claude mobile app guide covers what it does well. The only way that app touches your code is as a Remote Control window into a session running on your machine — the agent is still on the computer.

The honest limitations of Claude Code from mobile

Four frictions survive every route: a screen too small for diffs and file trees, a software keyboard that fights precise paths, fiddly permission prompts, and mobile networks that drop exactly midway through a long agent run.

Every route above works. None of them makes an iPhone a good development environment. The friction is real:

• The screen is tiny. Claude Code emits diffs, file trees, and long tool output. On a 6-inch screen you're scrolling constantly and losing context.
• The on-screen keyboard is the real killer. A coding agent lives on precise prompts, file paths, and the occasional Ctrl-C. Software keyboards are bad at all three. Most iOS terminal apps add a modifier row, which helps but doesn't fix it.
• Permission prompts are fiddly. Approving a tool call means hitting a small target while the agent waits. Doable, annoying.
• Mobile networks and long runs don't mix. A 15-minute agent run and a phone that switches from Wi-Fi to LTE in an elevator are natural enemies. This is exactly why tmux is mandatory on the SSH route, and why Remote Control's auto-reconnect matters on the official one.

What actually works: the verdict

Treat the phone as a remote control rather than a workstation. It is genuinely good for kicking off a known run, checking a long task from the sofa, approving a step, and typing a short course-correction — not for real development.

Treat Claude Code on iPhone as a remote control, not a workstation. It's genuinely great for:

• Kicking off a run you already know you want ("run the test suite and fix what's broken")
• Checking on a long task from the sofa or a queue
• Reading the output and approving or rejecting a step
• Typing a short course-correction when the agent goes sideways

This matches how it plays out for us. We run this whole site out of Claude Code, and the long agentic runs we kick off are exactly the kind of thing you want to glance at from the couch rather than babysit at the desk. The lesson that stuck was the boring one at the heart of the SSH route: a session pinned inside tmux is what lets you drop connection, walk away, and reattach mid-run without losing anything — treat Claude Code on iPhone as a way to peek at and nudge that persistent session, and it earns its place; ask it to be your editor, and it won't.

Claude Code on iPhone is bad for writing prompts longer than a sentence or two, reviewing a large diff properly, or anything resembling real development. The workflow that clicks for most people: start the work at your desk, keep it going from your phone, finish it at your desk. Understanding how Claude Code sessions persist makes that handoff much smoother, and if you want the same continuity on a bigger screen, Claude Code desktop covers that side. You can grab the official apps from Anthropic's download page.