Claude Code Remote Control: What It Is, How to Start It, and Which Mode to Pick

You start a refactor at your desk, then have to leave. The tests are still running and Claude is waiting on a decision. You want to check in from your phone without moving the work to someone else's server. That is the problem Claude Code Remote Control solves. It connects claude.ai/code, or the Claude app on iOS and Android, to a Claude Code session that keeps running on your own machine. You pick it up from a phone, a tablet or any browser.

This guide is based on the official documentation at https://code.claude.com/docs/en/remote-control. We explain it in our own words, so check the page for the latest details.

What Claude Code Remote Control is

Claude keeps running locally the whole time. Your filesystem, MCP servers, tools and project configuration stay on your computer, and typing @ in the remote window autocompletes paths from your local project. The web and mobile interfaces are only a window into that local session.

That has one consequence: your computer has to stay on, and the claude process has to keep running. If the laptop sleeps or the network drops, Claude Code reconnects on its own when the machine is back online.

What you get:

• The conversation, subagents and dynamic workflows stay in sync across the terminal, browser and phone, so you can send messages from any of them.
• You can attach photos or files from the Claude app or claude.ai/code. Photos reach Claude directly. Other files are downloaded to your machine and passed along as @ file references.

Requirements

Check these before you start:

• Plan: Pro, Max, Ultra, Team or Enterprise. API keys are not supported. On Team and Enterprise, an Owner must first enable the Remote Control toggle in Claude Code admin settings.
• Sign-in: run claude, then /login to sign in through claude.ai.
• Endpoint: it does not work with Amazon Bedrock, Google Cloud's Agent Platform or Microsoft Foundry. It also fails if ANTHROPIC_BASE_URL points at a host other than api.anthropic.com, or if you sign in through an enterprise Claude apps gateway.
• Feature flags: if CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC or CLAUDE_CODE_DISABLE_FEATURE_FLAGS is set, Remote Control is unavailable. DISABLE_TELEMETRY and DO_NOT_TRACK alone are allowed, unless your organization requires Trusted Devices. They need Claude Code v2.1.283 or later.
• Workspace trust: in an untrusted directory, claude remote-control asks Trust <directory>? [y/N]. If its input or output is not a terminal, it exits with a Workspace not trusted error.

Organizations with Zero Data Retention or the HIPAA configuration cannot enable it.

Start a session

There are several entry points. Pick the one that matches what you are doing.

Server mode runs a process that waits for remote connections:

claude remote-control

claude rc is an alias. The first time, it asks Enable Remote Control? (y/n). It then prints a URL, and you can press spacebar to show a QR code for your phone.

Interactive session gives you a normal terminal session that is also reachable remotely, so you can type locally and remotely:

claude --remote-control
claude --remote-control "My Project"

--rc works as a short form.

From an existing session carries over your conversation history:

/remote-control
/remote-control My Project

/rc is the short form. In the VS Code extension and the Claude Desktop app, /remote-control or /rc also works, but VS Code does not accept a name or show a QR code.

Server mode flags

These go after remote-control:

Flag	What it does
--name "My Project"	Sets the session title in the claude.ai/code list
--spawn <mode>	same-dir (shared directory), worktree (a git worktree per on-demand session), or session (exactly one session)
--capacity <N>	Maximum concurrent sessions; default is 32
--permission-mode <mode>	Starting permission mode, such as acceptEdits
--settings <file>	A JSON settings file path (v2.1.321 or later)
-c, --continue	Bring back the session the last server in this directory started with
--session-id <id>	Bring back one session by ID

Other flags include --verbose, --debug-file <path> and --[no-]create-session-in-dir. Run claude remote-control --help to print the full list. --sandbox and --no-sandbox were removed in v2.1.300 and now make the server exit with an error. To sandbox the sessions, turn on sandboxing in a settings file.

In same-dir or worktree mode, press w in the server terminal to toggle between them. The mode you switch to becomes the saved default for that project.

Connect from another device

Once a session is active, you have three options:

1. Open the URL Claude Code displays in any browser.
2. Scan the QR code to open it in the Claude app.
3. Open claude.ai/code or the Claude app and find the session by name. In the mobile app, tap Code. Online sessions show a computer icon with a green dot.

If you do not have the app yet, run /mobile inside Claude Code for a QR code that opens the right app store.

In an interactive session, an /rc active indicator links to the session. Run /remote-control again to see the URL, a QR code and a disconnect option.

Resume after stopping the server

Stopping claude remote-control with Ctrl+C makes its sessions stop responding. In the same directory, claude remote-control brings back every session, --continue brings back only the first one, and --session-id <id> brings back one by ID. This works for about four hours after the server stopped. For sessions started with claude --remote-control or /remote-control, resume with claude --continue or claude --resume.

Auto-connect and push notifications

Remote Control only starts when you ask for it. To connect every interactive session, run /config and set Enable Remote Control for all sessions, or set remoteControlAtStartup to true in ~/.claude/settings.json. A checked-in project settings file cannot turn it on, though it can turn it off with false.

For phone pushes, install the Claude app, sign in with the same account, allow notifications, then enable Push when Claude decides or Push when actions required in /config. If /config shows No mobile registered, open the app on your phone to refresh its push token.

Security

Your session makes outbound HTTPS requests only and never opens inbound ports. Traffic goes through the Anthropic API over TLS. While connected, the transcript is stored on Anthropic servers so devices stay in sync and the session can reconnect. Execution and file access stay local. The disableRemoteControl setting turns the feature off entirely.

Trusted Devices is an optional beta that requires an enrolled device and a sign-in no more than 18 hours old before you can view or steer sessions.

Remote Control or something else

Remote Control and cloud sessions share the claude.ai/code interface, but cloud sessions run on cloud infrastructure. Use Remote Control when you are mid-way through local work. Use a cloud session when you want no local setup, a repo you have not cloned, or parallel tasks.

The docs also compare Dispatch (message a task from the mobile app, runs on Desktop), Channels (push events from chat apps), Slack, self-hosted environments and scheduled tasks. Remote Control is the one for steering in-progress work from another device.

Common problems

• Command exits with an error: sign in with /login. Without an eligible login, claude remote-control exits, while claude --remote-control starts and shows a failure notification.
• Connection failed in an interactive session: run /remote-control to reconnect, unless the reason says another connection took over or the session was ended elsewhere.
• Not available at all: check the endpoint and feature-flag variables above.

Something to try in Claude Code

If you want a project to try in Claude Code, take a look at the free agent skill "Roller Diner Hero" from our open-source repository: a website hero on five seconds of generated footage, with a teddy-bear waitress in a mint uniform and paper hat roller-skating through a sunlit 1950s diner with red booths, chrome stools and a checkerboard floor, holding up a tray with cherry pie and a milkshake. A red script word and a giant cream "UP!" sit behind her, and she is cut out of every frame and drawn over them. Around it are a pill navigation, a guest-check card that gets stamped PAID, a starburst badge and a scrolling ticker. Your agent writes one hero.json with your words and links; the build checks every length and writes a page and an embed, and a recorder makes a 1080p MP4 with sound. Install it with npx skills add lookbkdev/lookbk-skills --skill roller-diner-hero, or with /plugin marketplace add lookbkdev/lookbk-skills and /plugin install roller-diner-hero@lookbk-skills; see the install guide, the skill page and the repository.

This post was drafted with AI assistance and checked against the skill's SKILL.md.