Classic server panels work the same way. Their backend holds an SSH key or runs an agent with root, and it can execute anything on your servers. If the panel's backend gets compromised, every server connected to it is compromised too.

We built Servor around the opposite rule. Your private key stays in your browser. Every command you send is signed there with Ed25519, and the agent on the server checks the signature before it runs anything. A command sent to 20 servers gets 20 signatures, one per server, so a signature can't be replayed on another machine.

Port 22 open to the internet means bots knocking all day. Behind NAT or a corporate firewall, it often isn't reachable at all.

The Servor agent works the other way around: it opens an outbound connection to us and keeps it alive. That gives you a terminal, files, cron, services and fleet commands without any inbound port, on Linux, macOS and Windows. The agent is a single compiled binary with no runtime to install, available for 7 OS/arch targets, and its source is public.

The typical setup we see in agencies managing 5 to 30 client VPS:

• an SSH client with shared keys

• an uptime monitor

• a status page tool

• a spreadsheet tracking who has access to what

• a note somewhere with the crons

Each tool has its own logins, its own bill, and its own way of drifting out of date.

"Run apt upgrade everywhere" sounds simple. In practice:

• a few servers are offline

• one hangs on a prompt

• another one has a different distro

In Servor, a fleet command fans out to every server you select. Each server gets its own signature, and results come back separately for each one, so you can tell straight away which servers failed and why. Commands you run often can be saved as recipes and replayed on any group.