How do I start Claude Code from my phone?

Run claude remote-control on a machine with your project, then open the Claude mobile app Code tab. Machines running Remote Control appear as device cards. Tap one, pick a folder, and the session starts on that computer with your local files and MCP servers intact.

Does Claude Code remote control survive laptop sleep?

Yes. As of August 2026, dropped connections recover automatically after brief sleep or Wi-Fi changes. Sessions time out after roughly ten minutes of sustained disconnect. Your phone and CLI stay synced on model and effort level.

Which Claude plans include Remote Control?

Remote Control is available on Pro, Max, Team, and Enterprise plans. Team and Enterprise require an admin to enable it. API-key-only access is not supported for Remote Control sessions.

Claude Code remote control now starts sessions from your phone

I have approved Claude Code permission prompts from a parking lot more times than I want to admit. Remote Control was supposed to fix that. Early builds felt brittle: stale sessions on my phone, reconnect loops, model drift between devices.

Anthropic's August 2026 update is the first version I would trust for a real handoff. You can start a session from your phone, auto-reconnect after sleep or Wi-Fi blips, and keep model, effort, and session state matched between CLI and mobile.

What actually changed

The Remote Control docs and Anthropic's dev thread spell out four reliability wins:

Feature	What it fixes
Phone-to-machine start	No SSH + terminal gymnastics on mobile
Auto-reconnect	Laptop lid close or Wi-Fi hop no longer kills the session
Live sync	Model + effort level match across phone and CLI
Clean shutdown	Phone shows offline within seconds when Claude Code exits

Slash commands also behave better on mobile now: /clear resets the phone view, /compact shows a compaction marker, and /diff opens the native diff sheet.

Diagram of phone connecting over HTTPS to a laptop running claude remote-control with local filesystem and MCP access

Three ways to turn it on

From the official guide:

Server mode (phone drives, terminal watches):

claude remote-control

The process stays in server mode, prints a session URL, and you can press spacebar for a QR code.

Interactive + remote (both active):

claude --remote-control

Mid-session toggle:

/remote-control

For always-on remote access, run /config inside Claude Code and enable Remote Control for all sessions, or set remoteControlAtStartup to true in ~/.claude/settings.json.

Managed settings can force Remote Control on org-wide. Project-level settings.json can turn it off for a repo, but cannot force it on for teammates.

Why local still matters

This is not cloud-hosted Claude Code. The session executes on your machine with your filesystem, MCP servers, and project config. The phone is a remote window, not a new environment.

That is the whole point for applied work:

Private repos stay local
Custom MCP tools (CRM, internal APIs) stay attached
.claude/ project settings travel with the repo

I wrote about adjacent patterns in ngrok + Cursor remote coding and Codex Windows computer use. Remote Control is Anthropic's first-party version of "keep the harness local, move the UI."

My workflow: couch to desk without a reset

Here is the loop I have been running on client repos:

Evening: claude remote-control on the dev laptop in the project root
Couch: open Claude mobile, tap the device card, pick the folder, kick off a refactor or test fix
Morning: resume at the desk in the same terminal session; phone stays on the live thread instead of archiving a ghost session

The quality-of-life win is not "code entirely on phone." It is no context rebuild when I change rooms.

Limits you should plan around

Constraint	Practical impact
Machine must stay on	Shutdown kills the session; sleep is OK
~10 minute disconnect timeout	Long flights need a fresh `claude remote-control`
No API-key auth	Billing goes through Claude subscription plans
Team/Enterprise gating	Admins must enable Remote Control org-wide

If you rely on scheduled Claude agents or cross-session messaging, Remote Control is complementary: it moves your interactive session, not background workers.

Security notes worth a one-minute read

Remote Control uses outbound HTTPS only. No inbound ports, no VPN required. Sessions are tied to your Claude account.

Still treat the phone like a second keyboard on your dev machine: lock the device, use org SSO where available, and do not leave claude remote-control running on a shared workstation logged into production credentials.

Checklist before you trust it on a client repo

Update CLI, desktop, and mobile to the latest builds (Anthropic ships these features cross-surface).
Confirm MCP servers reconnect after a forced sleep test.
Verify /diff on mobile for the file types you edit most (TSX, Python, infra).
Document for the team: Remote Control is not a replacement for CI or security plugins.

Bottom line

Remote Control went from demo to daily driver when Anthropic fixed reconnects and phone-initiated sessions. The interesting engineering is still local: your MCP graph, your repo, your agent discipline files.

If you are wiring agentic coding into ops workflows and want a second pair of eyes on harness design, book a free discovery call. I help teams ship agents that survive real devices, not just demo laptops.