Claude Code Remote Control: Run Your Session From Your Phone

The best part of Claude Code Remote Control is what doesn't change: your session still runs on your laptop with full access to your repo, your MCP servers, and your config. You're just attaching a phone or browser as a second screen. We kick off a long refactor at the desk and approve steps from the couch — same session, different device. Here's how to turn on Claude Code Remote Control, use it from your phone, and understand exactly what stays local.

What Claude Code Remote Control is

Remote Control links a remote front end — the Claude app for iOS and Android, or claude.ai/code in a browser — to a live Claude Code session on your computer. The key word is link: the agent doesn't relocate to the cloud. It keeps executing on your machine, and Claude Code Remote Control simply mirrors that session to your device so you can read output, answer prompts, and approve edits from anywhere.

That local-first design is why it's genuinely useful rather than a toy. Because the session stays on your hardware, you keep everything that makes Claude Code powerful:

How to enable Claude Code Remote Control

Turning it on takes about a minute. Claude Code Remote Control requires Claude Code v2.1.51 or later, so start by confirming your version:

claude --version

If you're behind, update Claude Code first. Then pick how you want to enable it:

For most individual users, running /remote-control inside an active session is the fastest path — it's the "let me grab this on my phone" button. If you always want the option, flip the /config toggle once and forget about it.

One note for teams: on Team and Enterprise plans, Claude Code Remote Control is off by default until an Owner enables the Remote Control toggle in Claude Code admin settings. That's a deliberate control so organizations decide whether local sessions can be reached from outside the machine.

Using Claude Code Remote Control from your phone

Once enabled, the flow is simple:

1. Start (or resume) a session on your machine and run /remote-control.
2. Open the Claude app or claude.ai/code on your phone, tablet, or another browser — use /mobile first if you still need to install the app.
3. Pick up the session from your device. You'll see the live conversation and can send prompts, answer permission requests, and approve or reject edits.
4. Walk away from your desk. The agent keeps working locally; you just monitor and steer remotely.

This turns dead time into progress. Kick off a large migration, then approve each planned step from your phone during a commute. It pairs especially well with plan mode: the agent writes a plan, and you approve it from wherever you are instead of being chained to the keyboard.

Is Claude Code Remote Control secure?

The security story is the strongest argument for the feature. Nothing moves to the cloud. When you start a Claude Code Remote Control session, Claude keeps running entirely on your machine — your source code, credentials, and tool access stay on your hardware. The remote device is a control surface, not a copy of your project.

A few practical points keep it safe:

• It's opt-in per session (/remote-control) or per config toggle — a session isn't reachable remotely unless you say so.
• Teams gate it centrally. Owners decide on Team and Enterprise plans, so remote access isn't silently available across an org.
• You still approve actions. Permission prompts and plan approvals work the same remotely, so nothing edits or runs without your say-so — the same guardrails as a local session.

If you're comparing this to fully cloud-based agents, the distinction matters: Remote Control gives you the reach of the cloud with the trust boundary of local execution. For the full feature reference, see Anthropic's Remote Control docs, and grab the app from the Claude download page.

Where Claude Code Remote Control actually earns its place

Remote Control isn't a novelty — it changes when and where you can make progress. A few workflows we lean on:

• Approve long runs on the move. Start a big refactor or a test-suite overhaul before you leave your desk, then approve each planned step from your phone on the train. The heavy lifting happens on your machine; you're just unblocking it.
• Babysit a slow, long-running task. Kick off a migration, step away, and glance at progress from the couch. When Claude hits a permission prompt, you clear it in seconds instead of losing an hour to a stalled run.
• Switch rooms without losing context. Move from your desk to a meeting and pick the same session up in a browser — no re-explaining, no lost history, because the session never left your machine.
• Pair from two places at once. Drive the same local session from a laptop and a phone: the big screen for reading diffs, the phone for quick yes/no approvals.

The common thread is that Claude Code Remote Control removes the requirement to be physically at the keyboard for work that's mostly waiting-and-approving — and that's a surprising share of agentic coding. The model plans, edits, and runs tests; you mostly say "yes, continue" at the right moments. Being able to say it from anywhere is the entire value.

It also composes cleanly with the rest of your setup. Because it's a normal local Claude Code session, you can resume it at your desk later with --continue as if you'd never left — Remote Control and local work are the same session, not two copies. And since execution stays local, your MCP servers and project tools keep working remotely with zero extra configuration, which is exactly what separates Claude Code Remote Control from a thin cloud chat window.