Running Claude Code on a Remote Server

Desk Terminal is a miniature computer styled after a 1984 all-in-one with a small colour screen. Its appearance does not establish it as an SSH client, so remote Claude Code access still requires a real supported computer and terminal.

Why run the agent on a server instead of your laptop?

A server can hold the repository, dependencies and running terminal process while a laptop supplies the SSH interface. If the laptop sleeps or its connection drops, a process inside a server-side terminal multiplexer can continue as long as the server itself remains available.

This arrangement is useful for long builds or a consistent development environment, but it transfers trust to the server. Patch the host, use least-privilege accounts, protect credentials, monitor usage and keep recoverable Git checkpoints. More uptime should not mean broader unattended authority.

How do you keep the session alive with tmux?

SSH to the server, change into the intended worktree and start a named tmux session before launching Claude Code. Detach from tmux rather than terminating the program. Later, list sessions and attach to the exact name.

tmux persistence depends on the remote host and its tmux server continuing to run. A reboot, terminated account process or provider failure can still end it. Claude's own conversation resume features and committed work provide separate recovery layers.

How do you reconnect from any machine?

Install a trusted SSH client on the new machine, authenticate with a protected key and verify the server's host identity. Attach to the named tmux session, then inspect the hostname, project path, branch and current prompt before entering anything.

Do not copy private keys into an ornamental display or expose SSH directly without normal hardening. Claude's current desktop application also documents SSH-based remote environments, while Remote Control is a different feature in which the Claude process keeps running locally and the web interface connects through outbound service traffic.

What does your desk still need locally?

It needs a supported device that can show the terminal, receive deliberate input and protect authentication material. A stable network, readable screen and ordinary keyboard matter more than local compute power. Approval decisions still require enough context to inspect diffs and command output.

The Desk Terminal can serve as a visual desk object or status companion only when software explicitly provides that function. Do not assume its small screen can host SSH or Claude Code. Keep actual remote access on documented hardware, and treat any ambient display as secondary to a full review surface. Test reconnecting once before depending on the setup during a long unattended build.