DEV Community: Vels Lobak

How to keep Claude Code running 24/7 (and use it from your phone)

Vels Lobak — Wed, 04 Nov 2026 20:54:43 +0000

Claude Code stops when your laptop sleeps. Here are 3 ways to keep it running around the clock and steer it from the Claude app on your phone.

I used to give Claude Code long tasks to work on, then I would switch to other chores and forget about it. Then I close the laptop and head out. Later, I would open the Claude app just to realize that it got interrupted midway or was stuck waiting for an approval nobody could give it...

If that sounds familiar, you're in good company. The Claude Code issue tracker is full of people asking the same question: how do I keep it going when I walk away from my desk?

The short answer: a local Claude Code session only runs while the computer it's on is awake and online. Remote Control lets you reach the session from your phone, but it doesn't keep it alive. So you either keep your laptop awake, or you run the session on a computer that never sleeps.

Below are the four ways I know how to make it work. Full disclosure up front: I built the last one. The others work fine, and for a lot of people they're the good enough, but I needed something more robust that is also easier to use and doesn't require babysitting.

Lets go trough them one by one.

First, the basics: Remote Control

Remote Control is Claude Code's built-in way to continue a session from another device. Inside a running session, type:

/remote-control

Then open the Claude app on your phone (or claude.ai/code) and continue the session.

Two things worth knowing:

The session still runs on your machine. If that machine goes to sleep, the session stops answering.
The link can drop without an error. If your phone shows the session but nothing updates, go back to the terminal and run /remote-control again.

Option 0: Claude Code cloud sessions (official, included in your plan)

Good if: your task is "take this repo and do X", and your code is on GitHub.

Since September 2026, Anthropic runs cloud sessions for Pro, Max and Team plans. You start a session from claude.ai/code, the Claude phone app, the desktop app, or run claude --cloud in your terminal. It runs on Anthropic's own machines, keeps going when you close your laptop, and there's no separate charge for the VM (it counts toward your normal usage limits).

The catch: it's a session, not a computer you can keep. After a while of inactivity the VM is reclaimed. This means that your conversation will continue where it stopped, but anything that was running in the background (a dev server, a long script, a scheduled loop) – all will be gone. And obviously you can't use their machine to host your app or website.

Its fine for one-off coding tasks. It's free with your plan.
But honestly - I tried it once and never used it again.

Option 1: Keep your own computer awake (free)

Good if: you have a desktop, or a laptop that can stay home and plugged in.

macOS: run caffeinate -dims in a terminal tab. The Mac stays awake while it runs. On a MacBook, closing the lid still puts it to sleep unless it's on power with an external display connected.
Linux: start Claude Code with systemd-inhibit --what=idle:sleep --why="claude running" claude.
Windows: Settings → System → Power → set sleep to "Never" when plugged in.

The catch: your laptop is now a server. It sits on your desk, on your home Wi-Fi, and Claude works on the same machine where your personal files and browser logins live. If the power or the Wi-Fi blinks, the session is gone until you're back to revive it.

Option 2: Rent a small server and use tmux (cheap, hands-on)

Good if: you're comfortable with Linux and SSH.

Rent a small Linux server. A few dollars a month is enough for light projects; but builds and tests need more memory.
Install Claude Code and sign in with your Claude account.
Start it inside tmux, so it keeps running after you disconnect: tmux new -s claude claude

In the claude session, type /remote-control. Detach with Ctrl+b, then d. Close your laptop. Open the Claude app on your phone or desktop and you will see this session appear in the list of sessions.

The catch: you're the sysadmin now. Updates, firewall, SSH keys, a full disk, Claude's login expiring, the process dying at 3am: all yours. tmux doesn't come back after a server reboot, and if Claude crashes, nothing restarts it. You find out about it the next time you check your phone.

Option 3: Give Claude Code its own computer (what I built)

This is the exact problem I built podway for. It gives Claude Code its own always-on computer in the cloud. We call it a pod: a private machine with your project, tools and services inside.

Your session is always running. Open it from the Claude app on your phone, no PC needed.
Nothing gets reclaimed. The pod stays on, so your dev servers, databases, workers and scheduled jobs keep running between sessions, and your files are still there tomorrow.
It recovers on its own. If Claude's login expires or the agent crashes, podway restarts it and reconnects Remote Control. Pod restarts and updates reconnect your session automatically.
It's the official CLI, signed in with your own Pro or Max plan. Podway doesn't proxy your model login or add any markup for tokens.
All your stuff stays on your personal computer. Claude works in an isolated computer, away from your personal files, browser sessions and home network.
You can always get in. There's a browser terminal for direct access to the cloud machine, and it works on your phone too.

There is one limitation, but it is kind of intentional: you can have only one claude session per pod. You always connect to the same session from all your devices and run your prompts in sequence. This is a genuine limitation from Anthropic, but it actually works to our advantage. It allows me to give each pod its own purpose – a focused target that it constantly pursues. This makes agent laser focused on the project and much more effective. Want to work on a new project? Separate your concerns, use another pod.

Pods on podway start at $4/month, and your first small pod is free during early access, no card needed.

So which one should you pick?

My suggestion:

One-off coding tasks on a GitHub repo → Option 0. It's included in your plan.
You have a computer that's always on anyway → Option 1. It's free.
You enjoy running and configuring servers → Option 2. Cheap, and you own every piece.
You want a machine that stays on, with your dev server, database and scheduled jobs still running tomorrow, without babysitting a server → Option 3.

If you try any of these and get stuck, tell me in the comments. Happy to help :)

Itzhak

Styled Components for React Native the way they should have been

Vels Lobak — Tue, 21 Dec 2021 17:00:00 +0000

To be honest, I've always felt that using 💅🏼styled-components to style View, Text, TouchableOpacity and other native components was very cumbersome.

React Native does not really work with DOM, so there is no real css engine underneath. This means that a lot of css functionality is simply not available and 90% of the time you will be using very basic css attributes: flex layout, colors, fonts, etc..

Then there is an issue with template strings - it is NOT easy to use them! Sometimes you end up creating nested template strings and needing to have special VS Code extensions to write css, and remember to put ; at the end of each line, etc..

And after you've managed all that, styled-components will generate the React Native styles from your template string and will pass them to the component. Yeah, cool...

So I thought to my self - why not use the native styles in the first place and avoid all the template string mess?

I googled a bit and found this this article, which inspired me to create an alternative to styled-components for React Native.

And thus the 💄styled-rn package was born :)

Why it better than 💅🏼styled-components/native?

Well, first of all, using styled components in React Native should be fun, but as I already mentioned, template strings are not fun.

Also:

styled-rn gives you access to ALL React Native style props
styled-rn is faster because it does not do tedious string template processing
styled-rn is easier to use (again, no messy string templates)
styled-rn is fully typed and has a nice API
styled-rn supports component attributes, custom props, theme via ThemeProvider, multiple style objects and more..
styled-rn has a shorter name ;)

Usage:

npm i styled-rn

The styled-rn API is almost one-to-one with the styled-components API, but everything happens in Javascript with full Typescript support:

import { styled } from "styled-rn";

export const Container = styled.View({
  flexDirection: "row",
  backgroundColor: "cyan",
});

Of course you can style any other component as well:

// Use with any component
export const CoolAndBoldComponent = styled(CoolComponent, {
  fontWeight: "bold",
});

You can also pass props to your components via the attrs key:

// Pass props to the styled component (attrs)
export const NonWrappingText = styled.Text({
  color: "blue",
}, {
  attrs: {
    numberOfLines: 1,
    ellipsizeMode: 'tail',
  }
});

I hope it is obvious to see how much cleaner and nicer the code looks compared to the template string approach.

💄styled-rn is a very small library, but it provides robust functionality and smart typings for all the styles, props and attributes.

You can find more examples and docs on the project's github

Thanks and Enjoy 💄styled-rn! :)