Remote Control: how it works

Claude Code runs locally; the web/mobile UI is a window into that session
The local process makes outbound HTTPS requests only and never opens inbound ports
When started, it registers with the Anthropic API and polls for work
When you connect, the server routes messages between your client and local session over a streaming connection (TLS)
Uses multiple short-lived, narrowly scoped credentials to limit blast radius

Security notes: https://code.claude.com/docs/en/security

Narration

Listen or continue with the transcript

Auto-play audio on each slide

Transcript

Under the hood, Remote Control is designed so your machine stays in charge.

When you enable it, Claude Code registers a session with Anthropic and then polls for updates. Your computer only makes outbound HTTPS requests; it does not open any inbound ports.

When you connect from the web or mobile app, Anthropic’s servers relay messages between your client and your local Claude Code session over a secure, streaming connection.

Remote Control uses multiple short-lived credentials with narrow scope. The goal is to reduce blast radius and make it easy to revoke access if needed.

Two practical gotchas to remember: your terminal window needs to stay open, and if your machine loses network connectivity for more than about ten minutes, the remote session will end and you’ll need to start it again.