Claude Code on your phone: continue a local task

A phone is useful for clarifying a task, answering the agent's question or checking a completed test. First, understand where the work runs: on your computer, on a remote server or in a cloud environment.

Remote Control connects a mobile or web interface to a local Claude Code session. Code and commands continue running on the computer hosting the agent. Your phone provides another interface to that session. Official overview.

• Official overview

What to prepare

Make sure Claude Code works in the intended folder and you are signed in with an eligible claude.ai account. Check Remote Control availability for your plan and your organization's policy: a regular API key does not replace the required sign-in. The computer must remain reachable and connected to the network.

Before leaving, give the agent a small deliverable and a stopping point. For example: identify a bug, prepare a fix and tests, then show the diff. That leaves you a concrete result to assess from your phone instead of an open-ended task.

Connect the current session

In an existing Claude Code session, enter:

/remote-control

Follow the connection instructions and open the session link on your phone using the same account. To start a new interactive session from the terminal:

claude --remote-control

You can open the link in a browser or use the Claude mobile app. Connection options are described in the Remote Control guide.

• Remote Control guide

Check the connection before a long task

Start with a harmless check that does not modify files:

State the current project folder and briefly describe the latest task. Do not change anything or start external operations.

Compare the response with your local terminal. Then send a short clarification from the computer and check that it appears on the phone. This catches cases where you opened a different project or session.

Do not share the session link or screenshots containing private paths publicly. Use a separate sample project for articles and demonstrations.

Give tasks from a small screen

A compact brief works well when continuing remotely:

Goal: fix the message shown when form submission fails.
Scope: only the form and its handler.
Checks: server error and successful submission.
Stop: show changes and results; do not publish.

This is a suggested template, not a required Claude Code format. Its purpose is to make the response easy to review. Ask for changed filenames and short explanations of debatable decisions so you can decide whether a full desktop review is needed.

Complex merges, migration reviews and large diffs are easier to inspect on a screen where you can examine the details. If a decision needs more context, ask the agent to preserve the current state and formulate the question.

What happens if the computer sleeps?

Remote Control does not transfer computation from your computer to the cloud. When the computer is unavailable, local work cannot keep running as it would on an always-on server. The client supports reconnection when connectivity returns, but you should still check the task's state. Connection behavior and limitations.

• Connection behavior and limitations

Before resuming, request a brief report: what is done, which processes are still running, which checks finished and which decisions are pending. This is more useful than resending the original task and risking duplicate work.

When SSH makes more sense

If your development environment already runs on a server, another option is a mobile SSH client and a persistent terminal session. You manage the remote machine and its environment yourself.

A private network such as Tailscale can connect your devices; installation is covered in its official quickstart. Network connectivity alone does not configure SSH, server users or terminal-session persistence.

• official quickstart

For a first attempt with local Claude Code, start with built-in Remote Control. A separate server makes sense when you actually need a machine that stays on independently of your laptop.