Web-based GPU monitoring dashboard for all your remote servers

Why gnvitop

Everything you need to
monitor your GPU cluster

🔌

SSH Auto-Discovery

Reads your ~/.ssh/config automatically. No extra configuration — if SSH works, gnvitop works.

Real-time Refresh

Dashboard auto-refreshes every 30 seconds with live GPU utilization, memory, and process data.

👤

Current User Highlight

Your processes are highlighted so you instantly see what's yours on a shared cluster.

🖥️

TUI Mode

Prefer the terminal? Run gnvitop --tui for a full-featured terminal UI — no browser needed.

🤖

Agent JSON Output

Use gnvitop --agent to get structured JSON of GPU availability — perfect for automation scripts.

🔴

Offline Detection

Servers that can't be reached are clearly shown as offline — no silent failures, no confusion.

Under the hood

How it works

⚙️
gnvitop
Python process
on your machine
🔐
SSH
Reads ~/.ssh/config
connects to each host
🖥️
nvidia-smi
Runs remotely
collects GPU stats
🌐
Browser
Live dashboard
auto-refreshes

See it in action

Beautiful dashboard,
zero configuration

gnvitop — localhost:5000
gnvitop dashboard screenshot

Get started

Up and running
in 30 seconds

Install
$ pip install gnvitop
Run
$ gnvitop
# Opens dashboard in your browser
Common flags
$ gnvitop -p 8080              # custom port
$ gnvitop --host 0.0.0.0       # expose to LAN
$ gnvitop --tui               # terminal UI
$ gnvitop --agent             # JSON output
Agent mode output
$ gnvitop --agent
[{"host":"server-a","gpu":0,
  "available":true,"util":12},
 {"host":"server-b","gpu":1,
  "available":false,"util":95}]