Skip to content
All guides

How-to

Claude Code Remote Control: set it up, start sessions from your phone, and keep them online

How Claude Code Remote Control works, the five ways to start it, how to start new sessions from the Claude app, what happens when your laptop sleeps or the Wi-Fi drops, and how to fix the common errors.

By 16 min read

Banner reading “Claude Code Remote Control: setup and limits”, with a laptop and a phone both showing the Claude Code mark, joined by a dashed line.

What is Claude Code Remote Control?

Remote Control connects the Claude app on iOS and Android, or claude.ai/code in a browser, to a Claude Code session running on your computer. Anthropic’s docs describe the phone and web as “a window into that local session”: the code runs, and the files stay, on your machine. Everything you set up locally still works, including your MCP servers, tools and project settings, and typing @ on the phone autocompletes file paths from your project.

It isn’t the same as Claude Code on the web. Cloud sessions run on Anthropic’s machines and keep going with your laptop closed. A Remote Control session runs on yours, so in the docs’ words “your computer has to stay on and the claude process has to keep running.” Anthropic’s own rule of thumb: use Remote Control when you’re in the middle of local work and want to keep going from another device, and a cloud session when you want to start a task without local setup or on a repo you haven’t cloned.

Remote Control launched in February 2026 as a research preview, and Anthropic’s Week 34 update (August 2026) took it out of preview. Older posts that call it a preview, or say it’s for Max only, are out of date.

What do you need to use Remote Control?

  • A Pro, Max, Ultra, Team or Enterprise plan. On Team and Enterprise, Remote Control is off until an Owner turns on the Remote Control toggle in Claude Code admin settings. Organizations with Zero Data Retention or the HIPAA configuration can’t enable it.
  • A claude.ai login, made with claude auth login or /login. API keys don’t work, and Remote Control is turned off when ANTHROPIC_API_KEY, ANTHROPIC_AUTH_TOKEN or apiKeyHelper is set, even if you’re also logged in. The long-lived token from claude setup-token can only make model requests, so it can’t start Remote Control either.
  • Claude Code talking to api.anthropic.com. It isn’t available through Amazon Bedrock, Google Cloud’s Agent Platform, Microsoft Foundry, an LLM gateway, or a custom ANTHROPIC_BASE_URL.
  • Feature flags left on. CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC, CLAUDE_CODE_DISABLE_FEATURE_FLAGS and DISABLE_GROWTHBOOK make it unavailable. DISABLE_TELEMETRY and DO_NOT_TRACK are fine on Claude Code v2.1.283 or later, unless your organization requires Trusted Devices.
  • A trusted project folder. The first time you run it in a folder, claude remote-control asks Trust <directory>? [y/N]. Run it from the project, not your home folder, where the answer is never saved.

If something on that list is wrong, claude doctor shows which Remote Control eligibility check fails.

How do I start Remote Control?

There are five places to start it. The one that matters most is the first: only a server can take new sessions you start from your phone.

WhereHowSessionsGood to know
Terminal, server modeclaude remote-controlUp to 32 at once (change with --capacity)No local prompt. Prints a URL; press the spacebar for a QR code. Alias: claude rc
Terminal, interactiveclaude --remote-control1A normal session you can type in locally and remotely. Alias: --rc
Inside a running session/remote-control1Carries the conversation so far over to your phone. Alias: /rc
VS Code extension/rc1No name argument and no QR code. Run it again to disconnect
Claude Desktop app/rc, or the session’s Remote Control switch1Find the session afterwards at claude.ai/code

The first time, Claude Code explains what Remote Control does and asks Enable Remote Control? (y/n). You answer once.

# Server mode: waits for connections and serves several sessions
cd ~/code/my-app
claude remote-control --name "my-app"

# Or an interactive session you can also reach from your phone
claude --remote-control "my-app"

To make every interactive session available without typing /rc, run /config and turn on Enable Remote Control for all sessions, or set "remoteControlAtStartup": true in your user ~/.claude/settings.json. A repository’s settings file can turn it off but not on, so nobody switches it on for you by committing a file. In the Desktop app the same option is Connect new sessions to Remote Control in Settings → Code.

Which server-mode flags are worth knowing?

FlagWhat it does
--name "my-app"Sets the session title shown in the app’s session list. Without it you get names like myhost-graceful-unicorn
--spawn worktreeGives each new session its own git worktree, so two sessions don’t edit the same files. The default, same-dir, shares one folder. Press w while it runs to switch
--spawn sessionServes exactly one session and refuses more
--capacity 8Maximum number of sessions at once. Default 32
--permission-mode acceptEditsThe permission mode new sessions start in (manual is an alias for default)
--settings ./rc.jsonApplies a settings file to the server’s sessions. Needs v2.1.321 or later
--continueBrings back the session the last server in this folder started with. Needs v2.1.200 or later
--chromeTurns Chrome integration on for the server’s sessions. Needs v2.1.273 or later

Put flags after remote-control, not before it. A global flag placed before the subcommand isn’t carried over to the sessions the server creates. Run claude remote-control --help for the full list.

How do I connect from my phone or another computer?

  1. Install the Claude app and sign in

    Use the same claude.ai account and organization you use in the terminal. If you don’t have the app yet, run /mobile in Claude Code for a QR code to the store. On an iPad, use the iOS app.

  2. Open the session

    Scan the QR code the server shows (press the spacebar), open the printed URL in any browser, or open the Claude app, tap Code and pick the session from the list. Remote Control sessions show a computer icon with a green dot when they’re online.

  3. Turn on notifications

    Run /config on your computer and turn on Push when actions required for permission prompts and questions, Push when Claude decides for updates such as a finished task, or both. If it says No mobile registered, open the Claude app once so it refreshes its push token.

If you don’t see a Code tab in the app, your plan or organization doesn’t include it. In an interactive session, an /rc active indicator in the terminal shows you’re connected; run /remote-control again to see the URL, show the QR code, or disconnect while the local session keeps running.

Can I start a new Claude Code session from my phone?

Yes, if the computer is running claude remote-control. People still ask for this on Reddit (one October thread is titled “Let us start sessions from Remote Control”), because the main Remote Control docs page doesn’t walk through it. Anthropic’s Week 34 update does: “Any machine running claude remote-control now shows up as a device card at the top of the Code tab in the Claude app. … Tap it to pick a directory and start a session there.”

  1. Start a server in the project

    claude remote-control --spawn worktree in the repository. Each session you start from the phone then gets its own git worktree instead of sharing your working folder.

  2. Tap the device card

    In the Claude app, open Code. Your computer appears as a card above the session list.

  3. Pick a directory and type the task

    The new session runs on your computer, like one you started at the keyboard.

Two limits. A server started with --spawn session serves exactly one session, so it can’t take new ones. And a single interactive session (--rc or /rc) can’t start others, although you can fork it from claude.ai/code with Fork in the session menu on v2.1.277 or later.

The docs don’t say which folders the directory picker offers. Replies in that Reddit thread say each server is tied to the directory you start it in, so people run one server per project, or start one in a parent folder and tell Claude which project to work in. Try it on your setup before you rely on it.

What can I do from the phone?

  • Chat with the session, and type @ to reference files in your project.
  • Attach photos and files. Claude sees photos directly; other files are downloaded to your computer and passed in as @ references.
  • See the diff. In a git repository, the diff pane shows your changes since the branch split from the default branch, including uncommitted edits.
  • Change the model or effort with /model sonnet or /effort high (pass the value as an argument).
  • See and stop the subagents and workflows the session is running.
  • Run /compact, /clear, /context, /usage, /rename and a few more. Terminal-only commands such as /plugin and /resume only work on the computer, and a line starting with ! is sent to Claude as text, not run as a shell command.

If a claude.ai connector such as Slack needs you to sign in again, do it on the computer: claude mcp login "claude.ai Slack" --no-browser in a terminal there (over SSH works), then /mcp reconnect claude.ai Slack in the session.

Can I approve or skip permission prompts from my phone?

You approve them from the phone. Permission prompts and questions from Claude wait until you answer them. Other dialogs Claude Code forwards close after five minutes and continue with their no-action default; the dialogExpiry setting changes that ("60s", "5m", "10m" or "never").

For fewer prompts, change the mode. In the app, tap + in the prompt box, then Permission, and choose Manual, Accept edits, Plan or Auto. You can’t choose Bypass permissions from the app. To start a server’s sessions in another mode, pass it on the computer:

claude remote-control --permission-mode acceptEdits

A Claude Code maintainer at Anthropic has suggested --permission-mode bypassPermissions for a headless server in issue #31908, but issue #29214 is still open with users who see prompts on the phone anyway. Only skip permissions in a sandbox or on a machine you’re prepared to lose.

What happens when my laptop sleeps, the Wi-Fi drops or I close the terminal?

This is where most “Remote Control stopped working” reports come from. Here is what Anthropic’s docs say happens in each case.

What happensResultWhat to do
The laptop sleeps or you close the lidNothing runs while it sleeps. Claude Code reconnects when the computer wakes upKeep it awake, or run the session on a computer that stays on (below)
The network drops, interactive sessionKeeps retrying, but can disconnect saying it “could not reach the Remote Control server for about 30 minutes”Run /remote-control to reconnect
The network drops, server modeGives up after roughly 10 minutes, and the claude remote-control process exitsRun claude remote-control again; on an always-on machine, restart it automatically
You switch Wi-Fi or a VPNRetries for up to three minutes if something answers with HTTP 403Check the VPN or proxy if it doesn’t come back
You close the terminal or the appThe session goes offlineBring it back (next section)
A session served by the server crashesThat session stops; the server keeps runningSend the session a message from the phone (v2.1.238 or later)
The claude.ai login expiresThe session stops making progress until you sign in againRun /login. Claude Code warns three days before
Your organization turns Remote Control offIt disconnects and doesn’t reconnect on its ownAsk your admin

The docs contradict themselves slightly on network outages: the Limitations list says an interactive session “retries for as long as the outage lasts,” while the next item, and the v2.1.232 release notes, describe a disconnect after about 30 minutes. Users also report in GitHub issues #95364 and #103005 that a Claude Desktop auto-update restarts the app and drops its Remote Control sessions. Anthropic’s Desktop docs say a restart for an update counts as quitting the app.

How do I bring a session back after it stops?

Stopping a server with Ctrl+C doesn’t delete its sessions; they just stop responding. In the same folder:

CommandBrings back
claude remote-controlEvery session the server was serving
claude remote-control --continueOnly the session the server started with
claude remote-control --session-id <id>One session. The ID is the part of its claude.ai/code URL after /code/
claude --continueA session you started with claude --remote-control or /remote-control

The server commands work for about four hours after the server stopped. They don’t work if you started the server with --no-create-session-in-dir, which archives its sessions when it stops. Since v2.1.323 (November 12, 2026), a server restarted in the same folder also keeps its computer identity for up to 30 days instead of showing up as a new computer.

If a second terminal says Remote Control not started here, run /remote-control there to move Remote Control to it.

Is Claude Code Remote Control secure?

Your computer only makes outbound HTTPS requests and never opens an inbound port, so there’s nothing to expose to the internet. It registers with the Anthropic API and waits for work, and messages between your phone and the session go through Anthropic’s servers over TLS, using several short-lived credentials that each do one job and expire separately.

One detail most guides leave out: while Remote Control is connected, the session transcript (your messages, Claude’s replies and tool activity) is stored on Anthropic’s servers so your devices stay in sync, under Anthropic’s data usage policy. The code still runs, and the files stay, on your computer.

Sessions belong to your claude.ai account. For more, turn on Trusted Devices (in beta): each device is enrolled, and a sign-in older than 18 hours has to be confirmed with Face ID, Touch ID, Windows Hello or a passkey. Admins can block Remote Control on a machine with the disableRemoteControl managed setting.

Why isn’t Remote Control working?

You seeFix
The session isn’t in the appCheck you started Remote Control (or turned on auto-connect) and that the app uses the same account and organization; /status shows which you’re signed in to
“Remote Control requires a claude.ai subscription”Run claude auth login and choose claude.ai. Remove ANTHROPIC_API_KEY or ANTHROPIC_AUTH_TOKEN wherever it’s set
“Remote Control requires a full-scope login token”You’re using a setup-token. Run claude auth login
“Remote Control isn’t enabled for this account”Run claude auth logout, then claude auth login. claude doctor shows which check fails
“Remote Control is disabled by your organization’s policy”An Owner hasn’t turned it on, the org uses ZDR or HIPAA, your role lacks the permission, or disableRemoteControl is set
“Remote Control is only available when using Claude via api.anthropic.com”Unset ANTHROPIC_BASE_URL or the Bedrock, Agent Platform or Foundry variable
“Remote credentials fetch failed”Re-run with --verbose. Remote Control needs outbound access to the Anthropic API on port 443
“Remote Control disconnected — Claude.ai login expired”Run /login, then /remote-control
The server won’t start from a script or a serviceRun claude remote-control once in a terminal as that user and answer the prompts, and pass --spawn=worktree or --spawn=same-dir

Anthropic’s troubleshooting section lists every message, and our 24/7 guide covers the errors you hit on a server.

How do I keep Remote Control online when my laptop sleeps?

Either keep the computer awake, or run the session on one that doesn’t sleep.

Keep the computer awake

This only works while it’s plugged in, and on most laptops only with the lid open.

WhereHow
Claude Desktop appTurn on Keep this computer awake for Remote Control in Settings → Code. Anthropic’s scheduled-tasks docs note that a similar Keep computer awake setting doesn’t stop the laptop sleeping when you close the lid
macOScaffeinate -i claude remote-control stops idle sleep, not lid sleep. With the lid closed, a MacBook stays awake only in clamshell mode: on power, with an external display
WindowsPower Options → Choose what closing the lid does → When I close the lid: Do nothing (Plugged in), and set sleep to Never while plugged in. Laptops with Modern Standby can still stop the session; powercfg /a shows whether yours has it
Linux (systemd)systemd-inhibit --what=sleep:idle:handle-lid-switch claude remote-control, which holds off sleep and the lid switch while it runs

The macOS and Windows steps come from Northflank and sotaaz; the Linux one adapts Northflank’s systemd-inhibit command to server mode. The Modern Standby problem is described in GitHub issue #71875.

Run it on a computer that stays on

OptionSurvives the lid closingSetup
A computer at home (Mac mini, desktop)YesRun claude remote-control in tmux
A VPSYesRemote Control inside tmux, restarted if it exits
Anthropic cloud sessionsYes, until they sit idleNo Remote Control needed; Claude Code only
A managed machine, such as ConcorsYesCreate a machine and start your agents

Anthropic recommends starting claude inside tmux or screen on a remote machine, so it keeps running after you disconnect from SSH. Two more things to plan for on a machine you don’t sit at: server mode exits after about 10 minutes without a network, so start it again automatically; and the claude.ai login expires, so renew it with /login when Claude Code warns you three days before.

Questions

Written by

Pierlorenzo Peruzzo

Engineer at Hologram, working on Concors

WebsiteGitHub

Sources

More guides

Let your agents
keep working.

Claude Code, Codex and OpenCode on a machine that stays on. Free on your own computer, no account needed.

Desktop app for macOS (Apple Silicon), Windows (x64) and