How to Use Claude Code From Your Phone: Every Option Compared
You can check on — and steer — a Claude Code session from your phone in at least four different ways. They differ in where the code runs, what happens when your laptop sleeps, and how much you can actually do on a small screen.
Nobody wants to write a feature on a phone. But checking whether the agent finished, answering the question it's been stuck on for an hour, or telling it "yes, go ahead and run the migration" — that's exactly what a phone is good for, and it's the difference between a task finishing while you're out and a task waiting for you to get home.
There are now several ways to do it. The important question isn't which app you use. It's where the code actually runs, because that decides what happens when you walk away.
The one question that sorts every option
For any setup, ask: if my laptop goes to sleep right now, does the work continue?
- If Claude Code runs on your laptop, the answer is no. Your phone is a remote control for a machine that has to stay awake.
- If Claude Code runs somewhere else — a server, a cloud environment — the answer is yes. Your phone is a window onto work that doesn't depend on you.
Keep that in mind through the list below.
Option 1: Anthropic's Remote Control
Remote Control, which Anthropic shipped as a research preview in February 2026, connects a Claude Code session running on your own machine to the Claude apps and claude.ai. You start a session in your terminal, hand it off, and then follow and direct it from your phone.
What's good: it's official, it's low-setup, and the interface is designed for a phone rather than being a terminal squeezed onto one. Your files stay on your machine; only the conversation travels.
The catch: the code still runs on your computer. That computer has to be on, awake, and online for the session to keep working. It's a great way to step away from your desk. It is not a way to close the laptop and go home.
Best for: stepping away for an hour while a desktop at home or in the office keeps running.
Option 2: Claude Code on the web
Anthropic also runs Claude Code in its own cloud environments, started from claude.ai or the Claude apps against a GitHub repository. You can launch a task from your phone, and it runs whether or not any of your devices is on.
What's good: truly device-independent. You can start work from your phone, not just check on it. For well-scoped tasks — fix this issue, add these tests, open a pull request — it's excellent.
The catch: these environments are built around a repository and a task, not a persistent machine. They suit work whose output is a pull request. They're less suited to work that needs long-lived state: a database with real data in it, a running server you want to click through, or an environment you've spent time setting up just so.
Best for: delegating discrete, reviewable changes.
Option 3: SSH into a server running tmux
The classic approach. Claude Code runs on a server inside tmux; you connect from your phone with an SSH client and reattach to the live session.
ssh dev@your-server
tmux attach -t agent
On iOS, Blink Shell and Termius are the usual choices. On Android, Termius, JuiceSSH, or Termux. Pair any of them with mosh if your connection drops a lot — it survives network changes that kill a plain SSH session.
What's good: full control, full terminal, runs regardless of your laptop, and the environment persists because it's your server. You're looking at exactly the session you left.
The catch: a terminal on a phone is a terminal on a phone. Reading output is fine; typing a long prompt is tedious, and scrolling back through tmux history with a touchscreen takes practice. You also have to build and maintain the server — we walk through what that involves in How to Run Claude Code on a Remote Server.
Best for: people comfortable running a server who want the real session, not a summary of it.
Option 4: a managed persistent workspace
A hosted environment where Claude Code runs in a long-lived workspace that belongs to one project, reachable from a browser, a phone, or SSH. It's option 3 without having to build the server yourself.
What's good: runs independently of every device you own, keeps state between sessions, and usually comes with the pieces around the agent that a raw server doesn't — a database, a URL to preview the running app, backups.
The catch: you're paying for managed infrastructure, and you get a managed environment rather than a machine you have root on.
Best for: ongoing projects where the agent works in the same environment for weeks.
Side by side
| Runs with laptop closed | Persistent environment | Setup effort | Phone experience | |
|---|---|---|---|---|
| Remote Control | No | Your laptop's | Low | Designed for it |
| Claude Code on the web | Yes | Per task | Low | Designed for it |
| SSH + tmux | Yes | Yes | High | Raw terminal |
| Managed workspace | Yes | Yes | Low | Browser or SSH |
Making the phone experience tolerable
Whichever route you pick, the same habits help:
- Write the prompt on a real keyboard. Start substantial work at your desk with a thorough brief. Use the phone to monitor and answer questions, not to author the task.
- Pre-approve routine commands. An agent that stops to ask whether it may run the test suite will spend your whole commute waiting. Set sensible permission rules up front — see Running Claude Code Unattended.
- Ask for summaries. "Summarise what you did and what's left" reads far better on a phone than 400 lines of scrollback.
- Keep a preview URL. Being able to tap a link and see the running app tells you more in five seconds than reading the diff.
- Push often. Whatever happens to the session, work that's on GitHub is safe.
The honest summary
If you mostly want to leave your desk for a bit, Remote Control is the simplest answer and it's free with your plan. If you want to hand off discrete tasks from anywhere, Claude Code on the web does that well. If you want an environment that keeps running and keeps its state — a database, a server, weeks of setup — you need the agent living on infrastructure that isn't in your bag, whether you build it or rent it.
EasySpawn runs Claude Code in a persistent workspace per project, reachable from a browser, your phone, or SSH — with managed Postgres, automatic SSL, and a live URL per branch so you can check the running app from anywhere. See how it works or join the waitlist.
Related: Agent Hosting Is Becoming Free. Here's What Isn't. · Why AI Coding Agents Need Persistent Workspaces
Keep reading
Running Claude Code Agents in Parallel With Git Worktrees
Two agents in one checkout will overwrite each other's work. Git worktrees give each Claude Code session its own files and branch on the same repository. How to set it up, and the parts nobody warns you about: ports, databases, and dependencies.
How to Write a CLAUDE.md That Actually Changes What Claude Does
Most CLAUDE.md files are either empty or a wall of generic advice Claude would have followed anyway. What to put in one, what to leave out, how the files load, and how to tell whether yours is working.