Claude Code Remote Control Can Start Sessions From Your Phone

Claude Code's Remote Control can now start a session from your phone. Anthropic's own release digest for Week 34 says that any machine running claude remote-control "now shows up as a device card at the top of the Code tab in the Claude app," and that you "tap it to pick a directory and start a session there." The same digest records that Remote Control is out of research preview, and as of September 28, 2026 the reference documentation carries no preview or beta label on the feature itself; only Trusted Devices, covered below, is still marked beta. (Claude Code, Week 34 - August 17-21, 2026; Claude Code Remote Control documentation)

The phone is a control surface, not the execution environment. Claude Code keeps running on the local machine, with local filesystem access, tools, MCP servers, and project configuration staying there while the conversation and progress are synchronized to the phone or browser. (Claude Code Remote Control documentation)

Start from the Claude app

The flow Anthropic documents is:

1. In a project directory on the machine, run claude remote-control.
2. Open the Claude app and tap Code.
3. Find the machine's device card at the top of the Code tab.
4. Tap the machine and pick a directory.
5. Start the Claude Code session on that machine, then continue from the phone.

The release digest describes the phone-start flow, while the reference documentation lists claude remote-control as server mode and explains that the session URL or QR code can open the session in claude.ai/code or the Claude app. (Claude Code, Week 34; Claude Code Remote Control documentation)

Server mode takes flags worth knowing before the first run:

Give these flags after remote-control, not before it. As of October 2026, the flag list has no sandbox switch: to sandbox the sessions the server starts, turn on sandboxing in a settings file. (Claude Code Remote Control documentation)

You can also expose an existing session instead of starting a new one. The documentation lists --remote-control and --rc for an interactive session, plus /remote-control or /rc inside an existing session. An existing session carries over its current conversation history when Remote Control is enabled. (Claude Code Remote Control documentation)

What stays on the computer

Remote Control does not move the project to the phone. Claude Code keeps code execution and filesystem access on the local machine, and the documentation says the local filesystem, MCP servers, tools, and project configuration are made available remotely through the connected session. (Claude Code Remote Control documentation)

That boundary makes the feature useful for long-running work without turning a phone into a development workstation. A session can keep running in a terminal or server process while you inspect progress, answer a question, send the next prompt, or attach a photo from the Claude app. A photo attached in the app goes to Claude directly as part of the message; Claude Code downloads other attached files to the machine and passes them to Claude as file references. (Claude Code Remote Control documentation)

The phone can also show subagents and workflows already running in the background. Stopping one from the phone stops that task on the machine, so treat the connected device as an active control surface rather than a read-only dashboard. (Claude Code Remote Control documentation)

Sync and reconnection improvements

The digest's phone-start entry sits beside the synchronization and reconnect behavior the reference documentation describes. As of September 28, 2026 it reads this way: the conversation and the progress of subagents and dynamic workflows "stay in sync across all connected devices, so you can send messages from your terminal, browser, and phone interchangeably"; switching conversations while a device is still connected does not archive the session; and if the local process stops, "the session goes offline until you bring it back." (Claude Code Remote Control documentation)

The documentation describes the reliability behavior separately: if the laptop sleeps or the network drops, Claude Code reconnects automatically when the machine comes back online. The local process still has to remain alive for the session to stay available. In an interactive session Claude Code keeps retrying for as long as an outage lasts; in server mode it gives up after roughly 10 minutes and the claude remote-control process exits, so you run the command again. (Claude Code Remote Control documentation)

This distinction matters: reconnect behavior helps an active session survive an interruption, while the phone-start feature is what lets you begin a new session from the phone on a machine where claude remote-control is already running. The two capabilities work together, but they solve different problems.

Plans and setup boundaries

The requirements section of Anthropic's documentation lists Remote Control for Pro, Max, Team, and Enterprise plans. Team and Enterprise workspaces have an extra administrative gate: it is off by default there until an Owner enables the Remote Control toggle in Claude Code admin settings. (Claude Code Remote Control documentation)

Remote Control does not use an Anthropic API key. The documentation says API-key authentication is unsupported and that the feature requires a claude.ai login. It is also unavailable when using Amazon Bedrock, Google Cloud's Agent Platform, or Microsoft Foundry, when you sign in through an enterprise Claude apps gateway, and when ANTHROPIC_BASE_URL points to a host other than api.anthropic.com. Anthropic's error reference adds that before v2.1.196, a custom ANTHROPIC_BASE_URL did not block Remote Control. (Claude Code Remote Control documentation; Claude Code error reference)

Before trying it, run Claude Code from the project directory at least once to accept the workspace trust dialog. The startup trust dialog never saves trust for your home directory, so start Remote Control from a project directory. Keep the machine awake and connected for the workflow you want to control remotely. If the process exits, the connected session goes offline rather than continuing to execute work on the phone. (Claude Code Remote Control documentation)

Two environment variables switch it off, and two no longer do

This is the requirement most likely to bite a privacy-conscious developer, and it changed with Claude Code v2.1.283, whose changelog entry says Anthropic "fixed Remote Control being unavailable on paid plans when telemetry is turned off" with DISABLE_TELEMETRY or DO_NOT_TRACK. Remote Control availability depends on feature-flag evaluation, and as of September 28, 2026 Anthropic's documentation says that whether Remote Control is available depends on which variable you set:

• CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC or DISABLE_GROWTHBOOK: Remote Control is unavailable. Set either one, in your shell or in the env block of a settings.json file, and Claude Code reports "Remote Control requires feature-flag evaluation" and names the variable it found. Unsetting the variable wherever it is set is the documented fix.
• DISABLE_TELEMETRY or DO_NOT_TRACK: Remote Control stays available unless your organization requires Trusted Devices, in which case you unset the variable. Using Remote Control with either variable set requires Claude Code v2.1.283 or later; from v2.1.154 through v2.1.282 either variable produced the same "requires feature-flag evaluation" error, so the fix on an older install is to update.

On versions before 2.1.154 the same configuration produced the more misleading "Remote Control is not yet enabled for your account" instead. The environment-variables reference lists the other features that need feature-flag fetching, among them /auto-mode-setup, claude import, and messaging sessions on other machines. (Claude Code Remote Control documentation; Claude Code environment variables; Claude Code changelog)

What an administrator controls

Three separate switches sit above the individual user, which matters if you are evaluating Remote Control for a team rather than for yourself.

The organization toggle. Off by default on Team and Enterprise. An Owner turns it on at Claude Code admin settings; it is a server-side organization setting.

disableRemoteControl. An IT administrator can turn Remote Control off on a device through managed settings, independent of the organization toggle and of how you are signed in. As of October 2026, the documentation says organizations with Zero Data Retention enabled, or with the HIPAA configuration applied, cannot enable Remote Control. In an organization with the HIPAA configuration, /status shows HIPAA on the Organization configuration line, and an Owner sees the Remote Control toggle grayed out in admin settings. (Claude Code Remote Control documentation)

Trusted Devices, currently in beta and off by default. As of September 28, 2026 the documentation lists it for Pro, Max, Team, and Enterprise: an Owner turns it on for a Team or Enterprise organization from Organization settings, Capabilities, Remote sessions, and a Pro or Max subscriber turns on Require trusted devices in their own settings on the Cowork or Account page. When it is on, viewing or steering a Remote Control session requires an enrolled device and a sign-in no more than 18 hours old; past that, the member confirms with Face ID, Touch ID, Windows Hello, or a passkey. Anthropic says biometric checks run on the device and that it never receives fingerprints, face data, or any other biometric information, storing only the device's public key and metadata such as display name, platform, and enrollment time. The setting applies to Remote Control only, not to Claude chat, the terminal, or the API, and it is not retroactive: sessions already running when the toggle is turned on continue without the device requirement until they end. For a lost or stolen device, the member removes it from the Trusted devices section of their account settings; if the member cannot sign in, an admin can use Sign out everywhere in the admin console to revoke every session and enrolled device for that member. (Claude Code Remote Control documentation)

A useful first workflow

Start with a bounded task on a machine you control: run claude remote-control in a repository, pick the project directory from the phone, and ask Claude Code to inspect or test a small change. Keep the first session visible on the machine so you can confirm permissions, MCP connections, and the project context before relying on phone-only control.

For longer work, use the phone to check progress and answer prompts, but keep the local execution boundary in mind. The session remains tied to the machine's filesystem and credentials, so anyone who can control the connected session should be treated as having meaningful access to that environment. One detail belongs in that assessment: Anthropic states that while Remote Control is connected, the session transcript, including your messages, Claude's responses, and tool activity, is stored on Anthropic servers to keep the conversation in sync across devices and to let the session reconnect after a network drop, retained under the Data usage policy. Execution and filesystem access stay on your machine, and the local session makes outbound HTTPS requests only, never opening inbound ports. The official documentation confirms the connection model and prerequisites; it does not make a general security guarantee for every project setup. (Claude Code Remote Control documentation)

Sources

Re-verified September 28, 2026.

• Claude Code, "Week 34 - August 17-21, 2026" release digest (releases v2.1.234 to v2.1.239), the Anthropic-owned source for the device-card phone-start flow and for Remote Control leaving research preview: https://code.claude.com/docs/en/whats-new/2026-w34
• Claude Code documentation, "Continue local sessions from any device with Remote Control," the load-bearing source for requirements, server-mode flags, the feature-flag environment variables, Trusted Devices, and transcript storage: https://code.claude.com/docs/en/remote-control
• Claude Code environment variables reference, "Features that need feature-flag fetching": https://code.claude.com/docs/en/env-vars
• Claude Code error reference, the v2.1.196 note on custom base URLs: https://code.claude.com/docs/en/errors
• Claude Code changelog, the v2.1.283 entry on DISABLE_TELEMETRY and DO_NOT_TRACK: https://code.claude.com/docs/en/changelog
• ClaudeDevs on X, "You can now start a Claude Code session directly from your phone," August 21, 2026, the original announcement. As of September 28, 2026 this URL answers with HTTP 200 and roughly 210KB of JavaScript; the post text is present inside that script as a bodyText field (between August 30 and September 27 the same fetch returned no post text at all). It is not load-bearing for any claim above; every claim rests on the Anthropic documentation: https://x.com/ClaudeDevs/status/2090933157243863142
• ClaudeDevs on X, "Your phone now stays in sync with what's running on your machine," August 21, 2026, same status: https://x.com/ClaudeDevs/status/2090933158795767846