Run Gemini CLI remotely with WebMux

1. Install and sign in to Gemini CLI

On the machine running WebMux, follow the official Gemini CLI docs (external link) to install it, and sign in as the same user that runs the WebMux service. WebMux does not install or sign in to CLIs and does not manage model keys.

2. Use the preset launch configuration

Interactive command

gemini

Preset arguments

--yolo

In “New session → Pre-type”, pick this launch configuration, choose a directory or worktree and create the session. You can also start it from a terminal on the server, detach with Ctrl+] and continue on your phone:

webmux run -- gemini --yolo

Upgrades never rewrite an existing launch-configs.json; if this CLI is missing, add it in “Settings → Launch configurations → New” with the command above. Presets use the “load shell startup files” wrapper so the CLI sees the PATH from ~/.bashrc or ~/.zshrc. Existing installs (Chinese)

3. Terminal, hybrid sessions and alerts

Terminal

Persistent PTY terminal: a browser disconnect does not end the process, and you can reconnect and replay the buffer; a service restart ends the session.

Hybrid session (chat ↔ terminal)

Not supported; it uses the terminal view. Offering an ACP interface alone does not mean WebMux's switching and resume behavior has been adapted.

“Needs you” alerts

Generic: depends on whether the CLI emits OSC 7501, Warp structured events, OSC 9 / OSC 777 notifications or BEL; WebMux documents no CLI-specific setup. Alert setup (Chinese)

4. Known limits

• Since 2026-06-18 Gemini CLI no longer serves free personal accounts or Google AI Pro/Ultra, so personal Google sign-in does not work; the preset is for paid Gemini API keys or Gemini Code Assist Standard/Enterprise users still within support.
• WebMux makes no promise about plans or quotas; personal-account users can migrate to Antigravity CLI per Google's notice and save their own launch configuration.
• The preset --yolo auto-approves tool calls; remove it in settings if you want to approve each step.