Claude Code Remote Control is useful from the phone or claude.ai — until the terminal on the machine dies. Close the window, crash, sign out, or reboot, and the remote client loses the session.
I built LiveClaude to treat those sessions like infrastructure instead of a leftover terminal window. I'm the author.
Remote Control (and Cursor My Machines workers) run as foreground processes. That is fine for a short interactive session. It is a bad fit if you expect the machine to stay reachable overnight or after a restart.
People usually invent one of:
tmux + a personal script
Those work. They also do not give you a multi-session UI, product-aware flags, or a place to complete the one-time trust/login prompts that need a real terminal.
LiveClaude is an MIT open-source supervisor + Avalonia desktop app, written entirely in C# / .NET (no Node, Python, tmux, or browser automation):
Same idea for Cursor My Machines workers when you want the same restart policy.
winget install LeandroCannizzaro.LiveClaude
curl -fsSL https://leandrocannizzaro.github.io/LiveClaude/install.sh | sh
curl -fsSL https://leandrocannizzaro.github.io/LiveClaude/install-macos.sh | sh
Repo: https://github.com/LeandroCannizzaro/LiveClaude
Product page: https://leandrocannizzaro.github.io/LiveClaude/
claude install stable path over a desktop-bundled CLI when you can
The interesting part is not "run a process forever." It is combining:
If you already solved this with a unit file and are happy, keep it. If you keep losing Remote Control after a reboot, LiveClaude might save you an afternoon.
Questions about the supervisor model or the VT path welcome in the comments.