Claude Code Remote Control Complete Guide — From Setup to Mobile Control

What Is Remote Control?

Claude Code Remote Control lets you control a Claude Code session running on your local machine from a mobile app or web browser. It was released as a research preview on February 25, 2026.

Here are the key points.

• Your code and development environment stays 100% local
• Mobile/web acts purely as a remote controller
• You can monitor tasks started on your desktop from the couch, or send instructions from a meeting room

graph TD
    subgraph LocalMachine["Local Machine"]
        A["Claude Code CLI"] --> B["Outbound HTTPS"]
    end
    subgraph AnthropicAPI["Anthropic API"]
        B --> C["TLS Streaming<br/>Message Routing"]
    end
    subgraph RemoteDevices["Remote Devices"]
        C --> D["Mobile App"]
        C --> E["Web Browser<br/>claude.ai/code"]
    end

Setup Methods

Method 1: Global Setting (Recommended)

To automatically enable Remote Control for all sessions, use the /config command.

# Inside a Claude Code session
/config

Select “Enable Remote Control for all sessions” from the settings menu. All subsequent Claude Code sessions will automatically have Remote Control enabled.

Method 2: Per-Session Activation

To enable Remote Control only for a specific session:

# Inside an active session
/remote-control

# Or the shorthand
/rc

Method 3: Start Directly from CLI

# Start a new Remote Control session
claude remote-control

# With detailed logs
claude remote-control --verbose

# Run in sandbox mode
claude remote-control --sandbox

Connecting from a Remote Device

Once Remote Control is active, you can connect in three ways.

QR Code Scan (Fastest)

Press the spacebar in the terminal to display a QR code. Scan it with the Claude mobile app for instant connection.

Copy Session URL

Copy the session URL displayed in the terminal and paste it into a browser on another device. It opens on claude.ai/code.

Select from Session List

Open claude.ai/code or the Claude mobile app to see your active sessions with a green status indicator.

Tip: If you run multiple sessions, use /rename to give each one a meaningful name. It makes finding the right session much easier on remote devices.

Network Architecture and Security

Remote Control’s security design follows these principles.

graph TD
    subgraph SecurityLayers["Security Layers"]
        A["Outbound Only<br/>No Inbound Ports"] ~~~ B["End-to-End Encryption<br/>TLS Streaming"]
        C["Short-Lived Credentials<br/>Independent Expiry"] ~~~ D["Permission Gates<br/>Approval Required"]
    end

Core Security Features

Feature	Description
Outbound Only	Only uses outgoing connections from local machine, no inbound ports opened
End-to-End Encryption	All traffic encrypted with TLS
Short-Lived Credentials	Each credential is scoped to a single purpose and expires independently
Anthropic Can’t See Your Code	Code stays local and never passes through Anthropic servers
Permission Gates	Every tool execution requires explicit approval

Session URL Security

Treat session URLs like passwords. Anyone with the URL can control your local session. Never share them unintentionally.

Practical Use Cases

Scenario 1: Monitoring Long Refactoring Jobs

Start a large refactoring task on your desktop and check progress from your phone while grabbing lunch.

# On your desktop
/rename "api-refactoring"
/rc

# → Scan QR code with your phone
# → Monitor progress in real time
# → Send additional instructions from mobile if needed

Scenario 2: Code Fixes During Meetings

When a bug report comes in during a meeting, send fix instructions without going back to your desk.

# On the mobile app, connected to your desktop session:
"The token validation logic in src/api/auth.ts is missing expiry checks. Please fix it."

Scenario 3: Test Suite Monitoring

Run your test suite and check results remotely, like a personal CI pipeline.

# Start tests on your desktop
"Run the full test suite and report only failures"

# → Move to the couch
# → Check test results on mobile
# → Send fix instructions for failing tests

Scenario 4: Multi-Project Management

Manage multiple project sessions by giving each a descriptive name.

# Name each project session
/rename "frontend-nextjs"
/rename "backend-api"
/rename "infra-terraform"

# On mobile: browse session list → select project → send instructions

CLI Integration Commands

Useful CLI options that work alongside Remote Control.

# Create a new web session
claude --remote "Fix the login bug"

# Transfer a web session to local terminal
claude --teleport

# Resume a previous session by name
claude --resume "api-refactoring"

# Show mobile app download QR code
/mobile

Limitations and Considerations

As a research preview, there are some limitations to be aware of.

Limitation	Details
Single Remote Connection	Only one remote connection per session
Terminal Must Stay Open	The claude remote-control process must keep running
Network Timeout	Session auto-terminates after ~10 minutes of network disconnection
Plan Restrictions	Max plan first, Pro plan rollout coming soon
Approval Required	Cannot be bypassed even with --dangerously-skip-permissions
Early Stability	Occasional API 500 errors may occur

Managing Session Defaults with /config

The /config command manages various settings beyond Remote Control. Here are the Remote Control-related options.

/config

# Among the displayed settings:
# ✅ Enable Remote Control for all sessions
#    → Auto-enables for every new session

Once configured, you no longer need to type /rc each time — Remote Control activates automatically when you start a session.

Community Alternative Solutions

Beyond the official Remote Control, community-built extensions offer additional capabilities.

Project	Features
claude-agent-server	WebSocket server wrapper, E2B sandbox deployment
claude-code-server	Web UI for CLI commands, Socket.io real-time streaming
247 Claude Code Remote	Tailscale VPN + Fly.io VM, access from anywhere

These projects aren’t officially supported but are useful when you need scheduling or programmatic access.

Conclusion

Claude Code Remote Control removes the constraint that you must be at your desk to code. Your code stays safely on your local machine while you can instruct and monitor your AI agent from anywhere.

Setup takes just one line with /config. While some instability is expected during the research preview phase, the core workflow is already practical enough for daily use.

Get started: Run /config in your terminal right now and enable “Enable Remote Control for all sessions.”