Remote VS Code on a Dedicated GPU Server

The best AI development experience is a local VS Code editor with remote execution on a dedicated GPU server. Your laptop stays cool, the GPU does the work, and the editor feels native. On our dedicated GPU hosting this is the recommended workflow for solo developers.

Contents

• Prerequisites
• Setup
• Useful extensions
• Performance tips

Prerequisites

• VS Code installed locally
• SSH key pair (ssh-keygen -t ed25519)
• Public key added to ~/.ssh/authorized_keys on the server
• Remote-SSH extension installed in VS Code

Setup

In VS Code: Cmd/Ctrl+Shift+P -> “Remote-SSH: Connect to Host” -> enter user@gigagpu-server.

First connect VS Code installs its remote-server component on the GPU box. Subsequent connects are fast.

For persistent config add to ~/.ssh/config on your laptop:

Host gigagpu
    HostName your-gpu-server.gigagpu.com
    User your-username
    IdentityFile ~/.ssh/id_ed25519
    ServerAliveInterval 60

Extensions

Install on the remote (not locally) for AI development:

• Python
• Pylance
• Jupyter
• Ruff (linting)
• GitLens

Remote-SSH handles extension split automatically when you install from the Remote window.

Tips

• Forward port 8000 to access vLLM from your browser: “Forward a Port” in the Ports panel
• Use VS Code’s Jupyter integration – no need for JupyterHub for solo work
• Keep your project on the remote filesystem, not synced – avoid drift
• Use tmux on the server so long-running jobs survive VS Code restarts

See JupyterHub for teams and SSH port forwarding.