How to use Claude Code from your phone

The first question is always the same: where does the agent actually run? Your phone is only a remote control. Either Claude Code runs on a computer you own (your laptop, a desktop, a server), or it runs on Anthropic's cloud. That choice decides what the agent can reach (your local files, databases, tools) and whether anything must stay switched on.

Details about Anthropic's products below come from their public docs as of September 2026. These features change often, so follow the links for the current state.

1. Remote Control (Anthropic): your machine, the Claude app

Remote Control connects the Claude app for iOS and Android (and claude.ai/code in a browser) to a Claude Code session running on your own computer. The phone is a window into that session; code execution and file access stay on your machine. Start it in your project folder:

claude remote-control          # server mode: waits for connections, press space for a QR code
claude --remote-control        # a normal interactive session you can also reach remotely
/remote-control                # inside a session that's already running

Scan the QR code, or open the Claude app, tap Code and pick the session. From the phone you can send messages (including photos), answer permission prompts, and look at a diff pane of your changes. Push notifications are opt-in: run /config in the terminal and enable "Push when Claude decides" and/or "Push when actions required".

Limits worth knowing: it needs a Pro, Max, Team or Enterprise plan and a claude.ai login (API keys are not supported, and on Team and Enterprise an admin must switch it on). The claude process must keep running, so if you close the terminal or the laptop sleeps, the session goes offline until the machine is back. Server mode (claude remote-control --spawn worktree) can host several sessions, each in its own worktree.

2. Cloud sessions (Anthropic): Anthropic's machine, no laptop needed

Claude Code on the web runs sessions on Anthropic-managed virtual machines. Start one from claude.ai/code, from the Code tab in the Claude mobile app, or from a terminal:

claude --cloud "Fix the flaky test in auth.spec.ts"

The session keeps running after you close your laptop, and you can steer it from any device. Each session has a diff view with inline comments and can create a pull request. Later, claude --teleport pulls a cloud session and its branch into your local terminal. It is available on Pro, Max and Team plans and some Enterprise seats.

The trade-off: the VM clones your repository from GitHub (or an uploaded bundle), not your working copy, so unpushed work, local databases, private network services and your own tooling aren't there unless you configure the cloud environment for them. Parallel cloud sessions share your account's usage limits.

3. SSH + tmux: the do-it-yourself route

The oldest approach, and it works with any CLI agent. Claude Code runs inside tmux on a computer you own; your phone connects over SSH and attaches to the same tmux session.

1. Make the machine reachable. Install Tailscale on the computer and the phone. Both join a private network, so you don't open any ports to the internet.
2. Turn on SSH. On a Mac: System Settings → General → Sharing → Remote Login. On Linux, install and enable openssh-server.
3. Start Claude Code inside tmux on the computer: tmux new -s claude, then claude.
4. Connect from the phone with an SSH app such as Blink Shell (iOS and iPadOS, supports mosh) or Termius (iOS and Android), using the machine's Tailscale name or IP.
5. Attach: tmux attach -t claude. When you're done, detach with Ctrl-b d; the session keeps running.

It's free apart from the SSH app, fully private, and you see exactly what you'd see at your desk. The downsides: a full-screen terminal on a phone is cramped, reading diffs means scrolling git diff output, and there are no notifications unless you script them. The computer must stay awake; see keeping Claude Code running with the lid closed.

4. Maestro Web: every workspace on your phone

Maestro Web is the free mobile companion to the Maestro desktop app. You sign in with GitHub, link your desktop once, and your workspaces appear on your phone as an installable web app: watch turns stream live, send, queue or steer messages, attach a screenshot, review diffs, and get Web Push notifications when an agent finishes or needs you.

Like Remote Control, the agents keep running on your own machine, so that machine has to stay on and awake. The difference is scope: Maestro Web shows every workspace at once, each in its own git worktree, and it isn't limited to Claude Code. Codex, Cursor, OpenCode, Kimi Code and Grok Build workspaces show up the same way. You choose which projects sync, and unlinking removes the mirrored data.

Side by side

Which one to pick

• You mostly work in one Claude Code session at your desk and want to keep an eye on it from the couch: Remote Control is built in and needs no setup beyond a command.
• You want work to continue with the laptop shut and your repo is on GitHub: cloud sessions, or move the agent to an always-on machine you own.
• You want full control and no third party in the loop: SSH + tmux over Tailscale.
• You run several agents in parallel, maybe from different vendors, and want one list on your phone showing which are working, which are waiting on you, and what changed: Maestro with Maestro Web. See Maestro as a Claude Code GUI and running Claude Code sessions in parallel.