A one-click application is an install script the server runs once, at its first boot, after the operating system has set itself up. The customer picks it on the create or reinstall form, fills in a short form, and the server's page shows the install's progress and, when it is done, how to log in. A recipe is a plain bash script run as root at the first boot, after the application.
Applications and recipes run on Linux images only: not on Windows or FreeBSD.
The built-in catalogue#
The panel ships 26 applications, all disabled until you enable them. Built-ins cannot be edited; you choose their images, plans and minimums, or Copy one into a custom application you can change.
| Category | Applications |
|---|---|
| Control panels | cPanel & WHM, Plesk Obsidian, DirectAdmin, HestiaCP, CloudPanel, CyberPanel, Webmin / Virtualmin GPL |
| Containers | Docker, Docker + Portainer CE, Coolify, Dokploy |
| Web stacks | WordPress, LAMP, LEMP, Node.js LTS |
| Databases | MariaDB server, PostgreSQL server (both on the loopback only) |
| Tools | Gitea, GitLab CE, Nextcloud, Nginx Proxy Manager, Uptime Kuma, n8n, WireGuard VPN (wg-easy), Mailcow |
| Games | Minecraft Java server (Paper) |
Each runs on the images its vendor supports (Debian, Ubuntu, AlmaLinux and Rocky Linux, depending on the application) and has a minimum size; control panels need 2 GB of memory or more, GitLab 4 GB, Mailcow 2 vCPUs and 6 GB. The card on the create form says how long an install takes when it is 15 minutes or more (cPanel: 20 minutes or more).
Applications that would let their first visitor become the owner (Portainer, Nginx Proxy Manager, n8n, wg-easy, Uptime Kuma, CloudPanel, Dokploy) are never opened to the world before the owner's account exists. An install is reported ready only once the application answers.
Licences#
- cPanel & WHM has no key field: cPanel licences are per IP address. WHM asks for the licence at the first sign-in and offers its 15-day trial there.
- Plesk: an optional key; empty starts Plesk's own 14-day trial.
- DirectAdmin requires a key (it has no trial); without one the create is refused.
Where an application can run#
| Application | Virtual machine | Container |
|---|---|---|
| Control panels, Dokploy, WireGuard, Mailcow | Yes | No: virtual machines only |
| Docker, Portainer, Coolify, Nginx Proxy Manager, Uptime Kuma, n8n | Yes | Only with a plan that has Nesting (Docker inside) on |
| The rest | Yes | Yes |
A container plan does not list what it cannot run. Mailcow also needs a public IPv4 address (not behind NAT) and outbound SMTP open for the server; it is refused otherwise, with the reason. The panel never changes a node's or server's SMTP rule for it.
Behind NAT#
An application that serves web pages declares its ports. A new server behind NAT gets a TCP port forward to each, named after the application ("WordPress HTTP"), counted against the plan's Port forwards included. When the allowance has no room, the create is refused with the number of forwards it needs: choose a plan with more port forwards. WireGuard's UDP port and Minecraft's port get forwards the same way. A reinstall does not add forwards: add them on the server's Network tab first.
Enable an application#
- Go to Images and ISOs › Images and open the Applications tab.
- Press Edit on an application.
- Under Images it runs on, tick the images to offer it with. An application with no image is not offered.
- Under Offered with, choose All plans or Chosen plans. Administrators creating a server with a size of their own may choose any enabled application.
- Check the Minimum size (vCPUs, memory, disk), save, and press Enable.
Customers also need the plan's One-click apps and recipes switch (Customer permissions).
What the customer does#
- On Create server (or Reinstall), under Install from, choose Application, then an application card. The cards are grouped by category, with a filter above more than eight.
- The image list narrows to the application's images, and its form appears. Password fields are never typed: the panel generates them and shows them on the server page once the install is ready.
- Optionally tick recipes under Run at first boot (optional) (up to 5).
The server's Overview then has an Application panel: waiting, installing (with each step as it is reported), ready or failed. When ready, it shows how to log in, with the passwords hidden behind Show. The owner also gets the "Application is ready" mail; passwords are in it only when the panel's mail settings include the root password. A failure names the step that was running; the log is /var/log/vmstack-first-boot.log on the server. After 6 hours without a report the panel says "no report from the server".
Custom applications#
- On the Applications tab press Add application, or Copy on a built-in.
- Application: Name, Slug (the API and billing modules name it so; it cannot be changed later), Description, Icon, Category.
- Offered on: images and plans, as above.
- Minimum size: Minimum vCPUs, Minimum memory (GB), Minimum disk (GB).
- Install: the Form (JSON, up to 12 fields of type text, password, email, select or checkbox), the Install template (a bash script), How to log in (a text with the same placeholders, shown on the server page and in the ready mail), and Web ports (up to four, forwarded behind NAT).
In the template and the login text, {{ name }} placeholders are replaced by the form's values as shell-quoted literals ('jane'). Write ADMIN={{ admin_user }} and use "$ADMIN" from there; a placeholder inside quotes is saved with a warning. Unknown placeholders are refused. The panel also fills in:
| Placeholder | Value |
|---|---|
{{ hostname }}, {{ fqdn }} |
The server's host name and full name |
{{ ipv4 }}, {{ ipv6 }} |
Its main IPv4 (the node's address behind NAT) and first IPv6 address |
{{ public_host }} |
The address the world reaches it at |
{{ web_url }} |
The URL of its first web port, through its forward behind NAT |
{{ ssh_command }} |
ssh root@host, or ssh -p PORT root@host behind NAT |
{{ server_id }} |
The server's id |
A password field is generated: letters and digits, 20 characters by default (16 to 64). Values are stored encrypted.
An example form with two fields:
[
{"name": "admin_user", "label": "Admin user name", "type": "text", "default": "admin", "pattern": "[a-z][a-z0-9_-]{2,31}"},
{"name": "admin_password", "label": "Admin password", "type": "password"}
]
and a template that uses them:
ADMIN={{ admin_user }}
PASS={{ admin_password }}
apt-get update && apt-get install -y nginx apache2-utils
htpasswd -bc /etc/nginx/.htpasswd "$ADMIN" "$PASS"
with How to log in set to Open {{ web_url }} and sign in as {{ admin_user }} with the password {{ admin_password }}.
Note: The script and its values are in the server's first-boot data, readable by root inside the server until the next reseed. Root is the customer, who sees these passwords anyway.
Recipes#
A recipe is a bash script a server runs as root at its first boot, after its application, in the order ticked. A recipe that ends with an exit code other than 0 stops the recipes after it and shows as failed.
- Images and ISOs › Images › Recipes is the panel's library (administrators): Name, Description, Script, and Who may use it: everyone, or administrators only.
- Servers › Recipes holds each account's own recipes.
Recipes take no form; the system values are in the environment (VMSTACK_PUBLIC_HOST, VMSTACK_WEB_URL, …). There is no "run now" on an existing server: recipes run at create and reinstall only.
Needs on the node#
Applications and recipes need an agent that runs a first-boot script. On an older node they are refused at create with the reason. The server needs outbound internet access to download the software it installs.