Short answer: yes. You can use Claude Code from a phone. But "from a phone" can mean three very different things:
- Your phone controls a Claude Code session running on your own computer.
- Your phone talks to a Claude Code session running in the cloud.
- Your phone is just an SSH terminal into a machine you run yourself.
Those are not interchangeable. The useful question is not whether the phone can display a prompt. It is where the code runs, which files and tools it can reach, and what happens when you close your laptop.
The fastest answer: use Remote Control for your current local project
Claude Code Remote Control lets you continue a local Claude Code session from the Claude app or from claude.ai/code in a phone browser. Claude Code still runs on your computer. Your local checkout, configured MCP servers, shell tools, and environment stay there; the phone is the control surface.
That makes it the right option when you are halfway through work on a repository that already lives on your Mac, PC, or Linux box. You do not need to push a branch, recreate local configuration, or expose a port to the internet. Remote Control uses outbound HTTPS from the machine, so there is no inbound port to open just to connect the phone.
Start a new session with:
claude --remote-control
Claude Code will show a session URL and QR code. Open or scan it on the phone, then sign in to the same eligible Claude account. If you already have a session open, use /remote-control (or /rc) inside that session instead. The official setup guide has the current eligibility and troubleshooting details.
The catch: the computer is still the computer
Remote Control does not turn an iPhone into the execution machine. If the local Claude Code process exits, the terminal closes, the computer sleeps, or its network disappears, there is no active local session for the phone to control. It can reconnect after the machine and session are available again, but it cannot keep a stopped local process working by itself.
That distinction matters most for long tasks. Remote Control is excellent for checking a build, answering a clarification, or steering an agent while you are away from your desk. It is not a substitute for an always-on environment when the real requirement is "keep working after my laptop lid closes."
Remote Control currently requires signing in with an eligible Claude subscription. The official documentation says it is available for Pro, Max, Team, and Enterprise users (with Team and Enterprise admin enablement where applicable); API-key authentication is not supported for this feature. Check the docs before setting it up, because account availability can change.
If you only have your phone: use Claude Code on the web
Claude Code on the web is a different model. The task runs in a cloud environment, not on the phone and not in your local terminal. You can start and manage those sessions from claude.ai/code or the Claude mobile app.
For a GitHub-backed project, Claude Code on the web works from a clone of the selected remote repository and branch. In practical terms: push the work you need before you start the cloud task. Uncommitted changes and local-only setup from your laptop are not automatically the cloud session's working state.
This is usually the cleanest route when you are genuinely phone-only and the work is already represented in a remote Git repository. It also avoids keeping your laptop awake. The trade-off is locality: the cloud session cannot simply inherit the local tools, filesystem paths, or MCP setup that you configured on your computer.
A simple way to choose
| What you need | Best starting point | Where Claude Code executes | Main trade-off |
|---|---|---|---|
| Continue work in the exact local checkout, with local tools and MCP servers | Remote Control | Your computer | The computer and session must remain available |
| Start or review GitHub-backed work while away from every computer | Claude Code on the web | A cloud environment | Work begins from the remote repository/branch, not your local state |
| Full control over an always-on machine | SSH to your own VPS or home machine | Your machine | You operate the machine, security, uptime, and session recovery |
The first two are official Claude Code workflows. The third is a general infrastructure pattern: use a phone SSH client to reach a server or home machine, then run Claude Code there. It can be powerful, but it also means you own the boring parts—patching the box, keeping credentials safe, and recovering a session after a reboot.
What "run it on my phone" does and does not mean
Can I run Claude Code directly on an iPhone?
With the official Remote Control workflow, no: Claude Code runs on your computer and the iPhone controls it. With Claude Code on the web, the work runs in a cloud environment. Both let you use a phone productively, but neither should be described as a full local iPhone development environment.
Does my laptop need to stay awake?
For Remote Control, yes—the local Claude Code process and the machine running it need to be available. For Claude Code on the web, no: the cloud session is separate from your laptop.
Will my local MCP servers work from my phone?
They can with Remote Control, because the execution stays in the same local Claude Code session where those servers are configured. A cloud session is a separate environment, so do not assume a local-only MCP server or shell integration will be present there.
Can I use Remote Control with an Anthropic API key?
Not currently. Remote Control uses Claude account sign-in and the supported subscription plans, not API-key authentication. This is easy to miss if you normally run Claude Code with a developer API key; confirm the current policy in the Remote Control documentation.
The deeper decision is persistence, not the screen size
If the job is occasional control—"I want to answer Claude from the train"—Remote Control is an unusually good fit. It keeps the work where you already configured it and makes the phone a safe, lightweight window into that session.
If the job is to create work from a phone with no machine waiting at home, Claude Code on the web is the more direct choice. Keep the branch pushed, be explicit about the task, and treat the cloud environment as its own workspace.
If the job is an agent that can keep building and operating an app while every personal device is asleep, you need an always-on execution environment. You can build and administer one yourself, or choose a hosted product. That is a different trade-off from Remote Control—not a hidden feature it provides.
VibeKit is in that last category, with an important boundary: it is not a mobile wrapper around your existing local Claude Code session. It provides a persistent agent for a hosted app that you can direct from a phone or CLI. That is useful when the outcome you want is an app that keeps moving without a laptop online; it is the wrong choice if you specifically need the files, shell, and MCP configuration in a local Claude Code checkout.
For the DIY versus hosted trade-offs in more detail, see Running a coding agent from your phone: three patterns and their tradeoffs. For a phone interface to an app-specific agent rather than a local developer session, see VibeKit's MCP server.
Before you leave the desk
Use this small checklist once and avoid most of the "why did it stop?" surprises:
- Remote Control: start the local session, confirm the phone can open it, and prevent the host from sleeping for work that must remain active.
- Cloud session: push the relevant branch first and assume local-only files, secrets, and integrations are not automatically available.
- DIY server: use a proper access path and session manager, then plan for reboots and updates before you hand it a long task.
- Any setup: make the agent's next action clear enough that you can inspect or redirect it from a small screen. A phone is a great control plane; it is a poor place to untangle ten terminal panes.
The practical answer is therefore: yes, you can use Claude Code from your phone today. Choose Remote Control when your computer is the workspace; choose a cloud session when the repository is the workspace.