How to Run Claude Code on Your Phone

You can run Claude Code on your phone right now, iPhone or Android. The fastest path: install Cosyra for iOS or Cosyra for Android, add your Anthropic API key, and type claude in the terminal. You get a persistent Ubuntu container with Claude Code pre-installed, no SSH tunnel to maintain, no laptop that needs to stay awake. 1 hour free on signup, no credit card.

This guide covers four ways to get Claude Code running on your phone, walks through setup step by step, and shows real workflows we use daily from a phone. If you already know you want the full picture of all four AI agents on mobile, start there, this page goes deep on Claude Code specifically. If you'd rather run OpenAI's agent, the Codex CLI on your phone guide walks through the same setup for Codex. The matching guides for Gemini CLI on your phone (free 60 rpm / 1000 rpd tier on Gemini 2.5 Pro) and OpenCode on your phone (open source, BYOK across providers) follow the same shape. If you have an iPad with a Magic Keyboard, the form factor changes the workflow enough that we wrote a separate Claude Code on iPad walkthrough covering Split View, Stage Manager, and dictation. For Android-specific setup (Termux's npm path, the Android 12+ phantom-process-killer fix, and foldable / DeX workflows on Pixel and Galaxy), see Claude Code on Android. And if the way you actually drive Claude Code is describing the change in plain English and letting it edit your repo, that's the workflow we walk through in vibe coding on your phone.

Why does Claude Code work on a phone?

Claude Code works on a phone because you direct the agent in short prompts instead of typing code character by character. The input is a sentence, the output is a diff you review and approve with a tap. That short-in, long-out ratio is what makes a glass keyboard viable for real engineering work, from CI fixes to multi-file refactors.

Claude Code is a terminal-based agent. You type a prompt, it reads your codebase, writes code across multiple files, runs tests, and proposes changes for you to accept or reject. The interaction model is short input, long output, you send a sentence, Claude Code sends back a diff. That ratio is what makes a phone screen workable. You're not typing if (err != nil) { return fmt.Errorf(...) } on a glass keyboard. You're typing "fix the nil pointer crash in the auth handler" and reviewing what the agent produces.

We think Claude Code on a phone is not a compromise, it's a better interface for certain tasks than a laptop. When you're on the train and a CI pipeline fails, pulling out a laptop means finding a seat, opening the lid, waiting for it to wake, SSHing in. On the phone, the container is already running. Tap, type the prompt, approve the fix, push. We've shipped production hotfixes from the back of an Uber this way.

What are the ways to run Claude Code on your phone?

There are four ways to run Claude Code on your phone as of 2026-04-17: SSH into a machine running Claude Code, Anthropic's Remote Control for an existing desktop session, Claude Code on the Web plus the Claude iOS app, or a cloud terminal like Cosyra with Claude Code pre-installed. Each approach solves a different constraint, so the right pick depends on whether you already own an always-on machine and which subscription you pay for.

1. SSH into a machine running Claude Code

The classic: run Claude Code on a desktop or VPS, SSH in from Blink Shell (iOS) or Termux (Android). Add Tailscale for zero-config networking and tmux to keep the session alive when you switch apps.

• Works when: you have a reliable always-on machine and you're comfortable with SSH + tmux setup.
• Breaks when: your MacBook sleeps, your VPS reboots, or your home network goes down. You're debugging infrastructure instead of writing code.
• Cost: free if you have existing hardware, plus whatever the VPS costs if you don't.

2. Claude Code Remote Control

Anthropic's built-in feature: run /rc inside a Claude Code session on your desktop, scan the QR code with your phone, and you get a mobile browser interface to the running session. You can read output and approve permission requests.

• Works when: you have an active Claude Code session on a desktop and you want to monitor or approve from your phone.
• Breaks when: you don't have a desktop session running, or you want to start new work from scratch on your phone. Remote Control is a viewer, not a full terminal.
• Cost: free (included with Claude Code).

If you're weighing Anthropic's Remote Control instead, we cover the tradeoffs in depth in Cosyra vs Claude Code Remote Control, the short version is that Remote Control is great when your laptop is always on, and Cosyra is the answer when it isn't.

3. Claude Code on the Web (plus the Claude iOS app)

Anthropic shipped a cloud surface of their own. Claude Code on the Web runs tasks on Anthropic-managed VMs at claude.ai/code, and the Claude iOS app lets you monitor and steer those sessions from a phone. You connect a GitHub repo, kick off a task (claude --remote "…"), and the VM clones, runs, and pushes a branch. Per Anthropic's own docs: 4 vCPU, 16 GB RAM, 30 GB disk, Ubuntu 24.04, and you can pull a session back into a local terminal with claude --teleport.

• Works when: your workflow is GitHub-native, you have a Claude Pro/Max/Team subscription (research preview eligibility, as of 2026-04-17), and you want Anthropic's own mobile surface.
• Breaks when: you want a persistent long-lived container, you work outside GitHub (GitLab/Bitbucket repos must be bundled and can't push back), you want OpenCode/Codex/Gemini alongside Claude in the same environment, or your organization has Zero Data Retention enabled (which disables cloud sessions).
• Cost: Claude Pro/Max/Team subscription required; no separate compute charge for the cloud VM. See Anthropic pricing.

Honest framing: if you already pay for Claude Pro and live inside GitHub, this is a very real option, and we'd tell you to try it. Where it falls short for us is the "I want a persistent Ubuntu container that also has Codex CLI, OpenCode, and Gemini CLI installed, reachable from an Android app too" story, which is why we kept building Cosyra.

4. Cloud terminal with Claude Code pre-installed

This is what we built with Cosyra. A native mobile app that connects to a persistent Ubuntu container in the cloud with Claude Code already set up. No SSH, no desktop dependency, no browser tabs that time out.

• Works when: you want to start or continue Claude Code work from your phone with zero setup. No other machine required.
• Breaks when: you have no internet connection (the container is in the cloud). Or if you need a specific corporate VPN that Cosyra doesn't route through.
• Cost: 1 hour free on signup, no credit card. Extend with a 10-hour, 7-day trial when you want more. After that, $29.99/month or $300/year. See pricing for details.

How do you set up Claude Code on iPhone or Android?

You set up Claude Code on a phone in about two minutes with Cosyra: install the app, sign in, paste an Anthropic API key into the container, and type claude. No Node install, no SSH tunnel, no desktop dependency. The three steps below walk through exactly what we run on a fresh install, with the actual terminal output we saw when we tested this on 2026-04-17.

This walkthrough uses Cosyra. Total time: about 2 minutes.

Step 1: Install and create a container

Download Cosyra from the App Store or Google Play. Sign up with email or Sign in with Apple/Google. The app provisions a fresh Ubuntu container on first launch. In our testing, that took about 15 seconds.

Step 2: Add your Anthropic API key

Claude Code needs an Anthropic API key. Get one from console.anthropic.com if you don't have one yet. Then set it in your container:

Your API key is stored in the container's .bashrc. Cosyra encrypts keys on-device before transmitting them to the container over an encrypted connection. You pay Anthropic directly at their standard token rates, Cosyra never touches your API billing.

Step 3: Clone a repo and start Claude Code

Claude Code indexes the codebase, identifies the issue, proposes a fix across the relevant files, and asks for your approval. You review the diff on your phone screen and accept or reject each change.

What do real Claude Code sessions from a phone look like?

Real Claude Code sessions from a phone look like short prompts, reviewed diffs, and git pushes from wherever you happen to be. Between meetings, waiting in a parking lot, or on the morning commute, the container stays running and you tap in to direct the agent. Below are three sessions we ran in the last month, captured verbatim, so you can see what fits into four to twenty minutes of phone time.

Abstract descriptions of "mobile coding" are useless without showing what the actual sessions look like. Here are three workflows we run regularly from a phone.

Fix a failing CI build between meetings

Slack notification: CI is red on main. You're between meetings with 8 minutes. Open Cosyra, pull latest, and point Claude Code at it:

Claude Code runs the tests, traces the failure to a missing null check in src/api/handlers.ts, proposes a two-line fix, and waits for your approval. You read the diff, approve, and Claude Code commits and pushes. Total time on phone: 4 minutes.

Start a feature while waiting for your kid's soccer practice to end

You've been thinking about adding rate limiting to the API. You have 20 minutes in a parking lot. Open the container, describe what you want:

Claude Code creates the middleware file, wires it into the Express app, writes tests, and runs them. You review each step. The container persists, so you can pick it up later on your laptop if you want, or keep going from the phone.

Review a PR on the morning commute

A teammate opened a PR overnight. You're on the subway with 12 minutes until your stop.

Claude Code reads every changed file, checks test coverage, and produces a structured review with specific line references. You post the review as a PR comment directly from the terminal using gh pr review --comment. Done before your stop.

What are the best tips for Claude Code on a phone?

The best tips for Claude Code on a phone are: wrap long tasks in tmux, write one detailed prompt instead of five vague ones, and use /compact when the context window gets heavy. Those three habits cut phone session friction more than any app setting. The full list below is the exact checklist we follow before starting any multi-step refactor from our own phones.

• Use tmux for long tasks. If Claude Code is working through a large refactor (10-20 minutes), run it inside tmux new -s work. Close the app, come back, reattach with tmux attach -t work. The agent keeps running.
• Be specific in your prompts. On a phone, you want fewer back-and-forth cycles. Instead of "improve the auth module," say "add JWT refresh token rotation to the auth module, update the tests, and make sure the existing session tests still pass." One detailed prompt beats five vague ones when you're typing with your thumbs.
• Use /compact to manage context. On long sessions, Claude Code's context window fills up. The /compact command summarizes the conversation and frees space. Useful when your phone session spans multiple tasks.
• Push from the container. The container has full git support. Set up your SSH key or use the GitHub CLI (gh auth login) and push directly. No need to switch devices. See Cosyra docs for SSH key setup.
• Session hibernation preserves state. Close the app and your terminal, files, and running processes pause. Come back hours later, everything resumes where you left off. No tmux reattach, no reconnecting.

How does Cosyra compare to other remote Claude Code options?

Cosyra compares favorably for persistence and device coverage, while Remote Control wins on price if you already have a desktop session, Claude Code on the Web wins if your life is on GitHub and you pay for Claude Pro, and SSH wins if you already run your own always-on box. The table below shows how the approaches line up on ten attributes as of 2026-04-17, so you can pick based on the trade-off that actually matters to you.

Remote Control wins on price and simplicity if you already have a desktop session running. SSH wins on flexibility if you have reliable infrastructure. Claude Code on the Web wins if you already pay for Claude Pro/Max and your life is on GitHub. Cosyra wins when you want a persistent Ubuntu container with multiple agents pre-installed, reachable from both iOS and Android apps, without depending on another machine or a specific Anthropic plan tier.

Frequently asked questions