Overview

You’re deep into a refactoring session with Claude Code at your desk and have to step away. Closing the terminal ends the session. Previously, this required an SSH tunnel or a third-party tool (happy, hapi, etc.). Now Claude Code has an official Remote Control feature. One command — claude remote-control — lets you resume the same session from a smartphone, tablet, or another computer.

How It Works

graph TD
    A["Local Machine <br/> claude remote-control"] -->|"HTTPS outbound only"| B["Anthropic API <br/> Message routing"]
    B --> C["claude.ai/code <br/> Browser"]
    B --> D["Claude Mobile App <br/> iOS/Android"]
    B --> E["Another Computer <br/> Browser"]
    C -->|"Real-time sync"| A
    D -->|"Real-time sync"| A
    E -->|"Real-time sync"| A

The key point: the session always runs on your local machine. Code never leaves for the cloud — your filesystem, MCP servers, and project settings remain intact. The local Claude Code process only sends HTTPS outbound requests; no inbound ports are opened. Anthropic’s API handles message routing in the middle.

If the network drops or your laptop sleeps, the session auto-reconnects when the machine comes back online — though a network outage longer than 10 minutes will time out the session.

Usage

Basic: Server Mode

claude remote-control

A session URL and QR code are printed in the terminal. Press Space to toggle the QR code so you can scan it with your phone.

Key Flags

Activating From an Existing Session

You can also activate Remote Control from an in-progress interactive session with /remote-control. Or go to /config and turn on “Enable Remote Control for all sessions” to apply it globally.

Three Ways to Connect

1. URL: Enter the session URL from the terminal directly in a browser
2. QR code: Press Space to show the QR code, then scan with your phone camera
3. Session list: Find the session by name in claude.ai/code or the Claude app (green dot = online)

Remote Control vs Claude Code on the Web

graph LR
    subgraph RC["Remote Control"]
        A1["Runs on local machine"] --> B1["Access to your filesystem"]
        A1 --> C1["Uses your MCP servers"]
        A1 --> D1["Preserves your project settings"]
    end
    subgraph Web["Claude Code on the Web"]
        A2["Runs on Anthropic cloud"] --> B2["Cloud VM environment"]
        A2 --> C2["No local config needed"]
        A2 --> D2["No repo clone needed"]
    end

Remote Control = “continue in my environment”. Web = “start fresh anywhere”.

Third-Party Alternatives

Community-mentioned third-party projects:

• slopus/happy, tiann/hapi — open-source tools with similar goals
• SSH tunnel to a remote terminal

The official Remote Control’s advantage: no separate server setup, TLS security via the Anthropic API by default. The downside, noted in community discussion, is that you have to set up the session in advance — which can feel less flexible than some open-source alternatives.

Limitations

• Plans: Pro, Max, Team, Enterprise (Team/Enterprise requires an admin to enable Claude Code first)
• No API key support: Authentication via claude.ai login only
• Terminal dependency: Closing the claude process ends the session
• Single remote connection: Outside server mode, only one remote connection per session is allowed
• Version: Requires Claude Code v2.1.51 or later (check with claude --version)

Insight

The real value of Remote Control isn’t “remote access” — it’s context preservation. A Claude Code session accumulates conversation history, the context of files already read, and active MCP server connections. Being able to switch devices without losing any of that is the point. A comment from the GeekNews discussion — “I can already see the YouTube videos about vibe coding from a café” — captures this feature’s use pattern perfectly. Combined with cmux’s notification system — monitoring multiple agents in cmux, then picking up with Remote Control on mobile when you step away — you have a complete multi-device agentic coding workflow.