01
Just type vio
A conversation, not a manual. Ask about your servers in plain language.
The tiny server agent.
vio runs, debugs and secures your Linux servers over SSH. One small binary, any model, and nothing changes without your approval.
$
curl -fsSL https://raw.githubusercontent.com/ThembinkosiThemba/vio/master/site/install.sh | sh
>
irm https://raw.githubusercontent.com/ThembinkosiThemba/vio/master/site/install.ps1 | iex
Then run vio. One binary, nothing installed on your servers.
$ viovio 0.1 · 2 hosts · qwen3 via ollama › is anything wrong with web-1? ● web-1 checking load, disk, services, firewall ● web-1 reading sshd config and auth log Two things need attention on web-1: 1. Password login is on, with 3,412 failed root logins today. 2. The firewall is off and Postgres :5432 is public.Want me to plan a fix? Nothing changes until you approve. › yes, fix both plan p-7f3a · 3 changes · lockout guard on sshd PasswordAuthentication no ufw allow 22, 80, 443 · deny 5432 · enable approve? [y/N] y ✓ applied · fresh SSH login ok · audit #1042›
Bring any model
OllamaAnthropicOpenAIOpenRoutervLLMLM Studio
Small tool. Serious guardrails.
Built for machines you can't afford to break.
Just type
02
Read-only by default
It looks before it touches. Every change is a plan you approve first.
03
Never locks you out
SSH and firewall changes roll back on their own unless a fresh login works.
04
Any model
A local model through Ollama, or any hosted API. Your keys, your choice.
05
Nothing on your servers
Agentless over SSH. One static binary, on your machine only.
06
Every action audited
A tamper-evident log of what it saw, proposed and did.
Point it at a server.
Read-only by default. Nothing changes without your approval.