Using Claude Code from your phone: every option compared
Use Claude Code from your phone with Remote Control, cloud sessions, Slack or a phone app for your Mac. Commands, requirements and which to pick.
The simplest way to use Claude Code from your phone is Remote Control: start it on your computer with claude remote-control (or /remote-control in a running session), then open the session in the Claude app or at claude.ai/code. Claude keeps running on your machine, with your files and tools; your phone is a window into it. If you want no computer involved at all, cloud sessions run on Anthropic's infrastructure instead.
#What are the ways to reach Claude Code from a phone?
They differ in where Claude runs and what starts the work:
| Option | Claude runs on | Good for |
|---|---|---|
| Remote Control | Your computer | Steering a session you already started |
| Cloud sessions (claude.ai/code) | Anthropic's cloud | Starting tasks with no local setup, repos you do not have cloned |
| Dispatch | Your computer (Claude Desktop) | Sending a task from the Claude mobile app |
| Slack | Anthropic's cloud | PRs and reviews from team chat |
| Channels | Your computer (CLI) | Reacting to external events such as chat messages or CI |
| SSH from a phone terminal | Your computer | Full terminal access, small keyboard |
| A phone app for your own setup | Your computer | Driving everything, not one session |
Most people want the first one, so the rest of this article spends its time there. Check the official docs for the current state of Dispatch, Channels and Slack, which are newer and still moving.
#How do I set up Remote Control?
You need a Pro, Max, Team or Enterprise plan with a claude.ai login (API keys are not supported), and an Owner must enable the toggle on Team and Enterprise. It does not work with Amazon Bedrock, Google Cloud's Agent Platform, Microsoft Foundry, or a custom ANTHROPIC_BASE_URL. Then pick the mode that fits:
# Server mode: stays running, can serve several sessions
claude remote-control # alias: claude rc
# Normal interactive session, also reachable remotely
claude --remote-control "My Project" # or --rc
or, from inside a session you already started:
/remote-control My Project
The last one carries your current conversation over. Each form shows a URL; open it, or press the spacebar in server mode to show a QR code and scan it. You can also open claude.ai/code or the Claude mobile app and find the session by name (in the mobile app, tap Code). Remote Control sessions show a computer icon with a green dot when online. If you do not have the app yet, /mobile shows a QR code for it. In an interactive session, an /rc active indicator shows the connection, and running /remote-control again opens a status panel where you can disconnect.
Useful server-mode flags:
--name "My Project"sets the title in the session list.--spawn worktreegives each on-demand session its own git worktree, so two phone-started sessions do not edit the same files. The defaultsame-dirshares the folder.--capacity <N>limits concurrent sessions.--permission-mode <mode>sets the starting permission mode.claude remote-control --continuebrings back the session after you stopped the server, for about four hours.
To connect every interactive session automatically, run /config and set Enable Remote Control for all sessions, or set remoteControlAtStartup to true in ~/.claude/settings.json. A checked-in project setting cannot turn it on for you, only off.
#What can I do from the phone, and what are the limits?
From the phone or browser you can send messages, attach photos and files, approve or deny tool calls, watch subagents, stop a background task, switch model or effort, and read the diff of your changes. The conversation stays in sync across terminal, browser and phone. In /config you can also enable push notifications to the phone for permission prompts and questions ("Push when actions required"), so you do not have to keep looking.
The limits are the ones you would guess:
- Your computer has to stay on. Remote Control is a window into a local process. If it sleeps or loses network, Claude Code reconnects when it is back, but nothing works meanwhile. Closing a laptop lid usually sleeps it.
- The
claudeprocess has to keep running. Stopping the terminal ends it. - Permission prompts still wait for you. That is the point, but it means an unattended session stops on the first one.
- Not everything works. A session started through
claude setup-tokenor an API key cannot use it, and the terminal-only parts (your shell, local servers) are not reachable from the phone.
#Is it safe?
By the docs: your local session makes outbound HTTPS requests only and opens no inbound ports. Traffic goes through the Anthropic API over TLS, using short-lived, narrowly scoped credentials. Execution and file access stay on your machine. The one thing to know: while connected, the session transcript, including your messages, Claude's responses and tool activity, is stored on Anthropic's servers so devices stay in sync. Organizations with Zero Data Retention cannot enable it. You can turn Remote Control off entirely with the disableRemoteControl setting, and Trusted Devices (beta) adds device enrollment and a recent sign-in requirement if you want it stricter.
#When should I use cloud sessions instead?
When the work does not depend on your machine: a repository you have not cloned, a task you want to run while your laptop is shut, several tasks in parallel. The trade is that nothing local is available: not your uncommitted changes, your local MCP servers or your tools. A rule of thumb: Remote Control for "I'm in the middle of this and want to keep going from the sofa", cloud for "start this and I'll check back".
#What about driving several sessions or agents from the phone?
Remote Control is per session: server mode can serve many, and the phone lists them, but it knows nothing about your projects, your task queue or your scheduled work. If you run many sessions across several repositories, you want an overview ("which of them are waiting for me?") more than a window into one. That is a different tool's job. Related reading: running several Claude Code sessions at once and scheduled Claude Code routines, which are the kind of work you then want to supervise from afar.
#A few habits that make it work
- Keep the computer awake. On macOS,
caffeinate -iin a terminal keeps it from idle-sleeping while it runs; a power setting does the same permanently. - Use worktrees for phone-started sessions (
--spawn worktree) so they never collide with what you are editing at your desk. - Pick a permission mode on purpose. Tapping Allow on a small screen is easy to do wrongly; read the diff first.
- Write the task like a ticket. Typing is slow on a phone. Put the context in CLAUDE.md and the task in one clear sentence. See how to prompt Claude Code.
- Watch your usage. Everything counts toward the same 5-hour window and weekly limit. See Claude Code usage limits.
#In Relais
Relais gives you both layers. Remote Control, the one described above, is on by default for the sessions Relais runs, so they show up in the Claude app and at claude.ai/code; idle sessions stay reachable longer, sessions come back after a restart, and you can turn it off per session. On top of that there is Relais Remote, an iPhone app, in beta: it drives the whole of Relais from your phone while your Mac is awake, with every group's sessions, approvals, the Board, agents' posts waiting for your yes, and notifications for a finished turn or a question. It needs an iPhone, a Mac that is awake and cloudflared (brew install cloudflared); you pair once with a QR code. Relais holds off idle sleep while a session works or waits for you. Details on the Remote page, and on agents for the posts you approve.