/servers

Real servers, sized to the hour.

Pick one of our locations, a size and an operating system. Grow CPU and memory whenever you like and come back down; give a server a different size at different times of day; back it up and snapshot it — without learning anyone else’s cloud.

Request early access Talk to us

11 locations
6 operating systems
Charged by the hour

Resize @vm-prod-1Standard · shared CPU

Now 2 vCPU · 4 GB · its disk is 80 GB

Grow the disk too

This is the size it runs now. Pick another to see what a resize would do.

The server is powered off for the resize and started again; billing moves to the new size the second it is confirmed.

11
locations
New York, San Francisco, Atlanta, Kansas City and more
6
operating systems
long-supported releases only
15s
between setup checks
called ready within seconds of it
10min
between money checks
every machine matched to its record

/what-a-server-does

Servers that behave the way you’d expect.

Everything here works today. The sizes, their prices and which locations offer them are shown on the create screen before you order anything.

Our locations, our sizes, our price.

Choose from 11 locations — New York, San Francisco, Atlanta, Kansas City, Toronto, Frankfurt, London, Amsterdam, Bangalore, Singapore, Sydney — and four kinds of size: Standard (shared CPU, the best value), Compute (dedicated CPU for CPU-bound work), Balanced (dedicated, a good default for databases) and Memory (dedicated, a lot of RAM). The price is quoted per month and charged for the hours the server exists, never more than the monthly price.

One place per city, whatever runs behind it — adding capacity changes nothing you see.
Billing starts when the machine is confirmed running and stops when it is confirmed gone. A server that never came up is never charged.
A stopped server still costs the same — its disk and address stay reserved. Delete it to stop paying.

New server · review

Location          EU Central · FrankfurtSize              Standard · 2 vCPU · 4 GB · 80 GB diskOperating system  Ubuntu 24.04 LTS (supported until 2029)Software          Docker · Nginx · Certbot · Fail2banSSH keys          alex@laptop, ci-deployBackups           weekly, from the start # the price of this size is shown here — per month, charged by the hourCreate server

Grow and shrink — the disk only when you say so.

Start small and move CPU and memory up or down as often as you like: those resizes can be undone. Growing the disk is a separate, permanent step you take only when you need the space — the screen says so before you confirm. A size whose disk is smaller than the server’s is shown as unavailable, with the reason.

Every size sorted into bigger, smaller and not possible — each saying whether it can be undone.
The server is powered off for the resize and started again; a failed resize brings it back up and says why.
The invoice shows both parts of the month: the old size’s hours and the new one’s.

14:00 · now running

8 vCPU · 16 GB

00–072 vCPU · 4 GB
07–198 vCPU · 16 GB
19–244 vCPU · 8 GB

A size for each time of day.

Give one server a bigger size during office hours and a smaller one at night and at weekends — up to 12 rules, in its own time zone. Rules are checked when you save them, so a schedule can only move between sizes that resize into each other: it can never make a permanent change at 3 a.m.

Weekdays 08:00 → a bigger size, 19:00 → a smaller one, weekends smaller still.
Every run is on the server’s activity; a failure turns the schedule off and tells you.
Other clouds add identical machines behind a balancer. This changes the size of the one you have.

vm-prod-1 · schedule · Europe/Berlin

# a different size at different times — same disk, so every step can be undoneMon–Fri  08:00  →  Balanced 4 vCPU · 16 GBMon–Fri  19:00  →  Standard 2 vCPU · 4 GBSat      00:00  →  Standard 1 vCPU · 2 GB next: Mon 08:00 · resize queued, then back to running

Operating system and software, chosen at create.

Pick Ubuntu 24.04, Ubuntu 26.04, Ubuntu 22.04, Debian 13, Rocky Linux 9, AlmaLinux 9 — long-supported releases only — and the software the machine itself should run: Docker, Nginx, Caddy, Certbot, Fail2ban, Git. Each comes from its maker’s own signed repository, so it is the latest stable release on the day the server is made. Setup is looked at every 15 seconds and shown step by step, with the version each installed.

Every server gets the baseline: sign-in with ssh keys only; a firewall that lets in only ssh, http and https; security updates installed automatically; swap on machines with less than 2 gb of memory.
A step that fails is named, and Try again reruns only the steps that did not finish.
A deploy asked for while setup runs waits, and starts on its own when the server is ready.

vm-prod-1 · Ubuntu 24.04 LTSSet up in 2 min 34 s

Swap2 GB
Security baselinekeys only · firewall · updates
Docker29.8.1
Nginx1.30.5
Certbot5.8.0
Fail2banwatching sshd
Gitinstalled

Server vm-prod-1 is ready — a deploy that was waiting for it starts now.

Looked at every 15 seconds while it runs

Backups and snapshots, restorable from the page.

Switch backups on at create or later: weekly (the last 4 weekly backups are kept), daily (the last 7 daily backups are kept). They are priced as a share of the server’s own price — 20% weekly, 30% daily. Snapshots are taken whenever you want while the server keeps running; one can restore the server, start a new one, and outlives the server it came from.

Restore from any backup the server has — its name typed back before the disk is replaced.
A new server from a snapshot runs what the snapshot’s disk runs: its operating system and software.
Snapshots are billed on the space they occupy, by the hour, and reconciled like everything else.

vm-prod-1 · backups and snapshots

Backups   weekly · Sunday, 00:00–04:00 UTC · the last 4 kept  Sep 21  02:14   ready    Restore…  Sep 14  02:09   ready    Restore… Snapshots taken while the server runs  before-upgrade   21.4 GB   ready   Restore · New server  pre-migration    19.8 GB   ready   Restore · New server # a snapshot outlives its server; restoring asks for the name typed back

A terminal in the browser — and your own SSH keys.

Open a shell on any running server from its page, in a tab of its own. It goes through the platform with a single-use ticket, closes after 30 idle minutes or 8 hours, and is recorded — who, when, from where, for how long. Add your organization’s public keys and choose which go onto a new server; we never hold your private key.

Opening a terminal is its own permission: a role can manage servers without logging in to them.
Sign-in with keys only — no passwords, no root password — on every server.
Every session is on the server’s activity and in the audit log.

vm-prod-1 · terminal

Connected to vm-prod-1 · session recordeddocker ps --format "{{.Names}}  {{.Status}}"crm-api     Up 2 hours (healthy)crm-worker  Up 2 hourscrm-mysql   Up 6 days (healthy)df -h //dev/vda1   78G   21G   57G  27% /

Nothing without a record.

Every machine carries the tag of the record that owns it, every call that can cost money is written down before it is made, and every 10 minutes both sides are compared. A machine nobody owns is found and removed; a record whose machine has vanished stops being billed that second, and you are told.

A create whose answer was lost is found by its tag — never made twice.
Billing follows the confirmed machine, not the request.
A delete that will not finish is retried until it does, and the server page says so.

Works with

Ubuntu
Debian
Rocky Linux
AlmaLinux
Docker
Nginx
Caddy
Certbot
Git

Every integration →

/faq

Servers, answered plainly.

Something else? Write to [email protected] or use the contact page — a person answers.

Whose hardware is it?

Ours to run and to answer for. Locations, sizes and prices are ours; which machines serve a location is our concern, and adding more changes nothing you see. Your servers never share a network with another organization’s.

How much does a server cost?

Each size has its own monthly price, shown on the create screen before you order. You are charged for the hours the server exists — the monthly price divided by 672 an hour — and never more than the monthly price in a month, whatever its length.

Can I make a server smaller again?

Yes, as long as the disk stays the same: CPU and memory move both ways as often as you like. The disk only grows, and only when you ask — because a disk can never be made smaller, a size with a smaller disk is refused with that reason.

Which operating systems are there?

Ubuntu 24.04 LTS (supported until 2029), Ubuntu 26.04 LTS (supported until 2031), Ubuntu 22.04 LTS (supported until 2027), Debian 13 (supported until 2030), Rocky Linux 9 (supported until 2032), AlmaLinux 9 (supported until 2032). Rocky Linux 10 and AlmaLinux 10 are not offered: Docker cannot start on their kernel with the container firewall the platform uses.

Can I bring my own server?

Yes — connect a server you already have for a small platform fee a month (it is in the price book), or create one in our locations.

Do I get root?

Yes. Add your SSH keys and they are installed for root on the servers you choose; the server page shows the command to copy. The browser terminal works too, and every session is recorded.

/start

Your first server, in minutes.

Request early access — a server is charged for the hours it runs, never more than its monthly price.

Request early access Talk to us