Quickstart: your first cluster in ~10 minutes
From an empty account to kubectl get nodes against your own cluster takes about 10–15 minutes,
most of which is waiting for the control plane to provision. You’ll need a
Hetzner Cloud project you can create API tokens in, and a payment
card.
-
Sign up
Create an account at console.paasbox.com — email or GitHub. You land in a team; everything you create (connections, servers, clusters) belongs to the team, so colleagues you invite later see the same things.
-
Add a payment method
Put a card on file. Nothing billable can start without one, and nothing is charged until you start a cluster — a one-time €5 signup credit covers your first hours of experimenting. See billing & spend controls for how the meter works.
-
Connect your Hetzner project
In your Hetzner Cloud Console, create an API token under Security → API tokens with read & write permissions, and paste it into the PaaSbox console. PaaSbox validates the token by listing your project, stores it encrypted, and never displays it again.
Read/write is required because managed node pools create (and delete) servers in your project — Hetzner tokens can’t exclude just the delete permission. The FAQ explains this in more detail, and adopted servers are protected by a separate never-delete invariant regardless of token scope.
-
Compose your cluster
Hit Create cluster and fill in the short form:
- Name — anything you like.
- Region —
nbg1(Nuremberg) is live today; other regions show “Register interest” (as of 2026-07-09). - Kubernetes version — take the default (currently a recent 1.35.x (as of 2026-07-09)).
- Node pool — one Managed pool with a
cpx32(4 vCPU / 8 GB) server type and 1–3 nodes is a good first shape. If you have grandfathered servers to adopt, you can add a Pooled pool now or later.
Everything else has working defaults — ~90% of clusters never touch an advanced setting.
-
Start it and watch it provision
Press Start. This opens the cluster’s subscription (prorated) and begins provisioning: the control plane comes up as managed pods on our platform, a small node image (~0.25 GB snapshot) lands in your Hetzner project, and your worker nodes boot and join over outbound-only connections — PaaSbox never needs inbound access to your project. The status moves from Provisioning to Ready in roughly 10–15 minutes.
Curious what exactly shows up in your Hetzner project? See from token to cluster.
-
Download your kubeconfig
On the cluster page, download the admin kubeconfig. It’s a real cluster-admin credential, short-lived (up to 8 hours), and issuing one is recorded in your activity feed. Then:
Terminal window export KUBECONFIG=~/Downloads/my-cluster.kubeconfig.yamlkubectl get nodes -o wideYour worker nodes report
Ready, running in your own Hetzner project. -
Deploy something
From here it’s ordinary Kubernetes —
kubectl, Helm, or Flux, your choice. The deploy your first app guide walks through a first deployment end to end.
Where to go next
Section titled “Where to go next”- Adopt your grandfathered servers — the reason PaaSbox exists: bring your pre-increase Hetzner servers in as worker nodes that are never deleted.
- Node pools — Managed vs Pooled, scaling, and how the two mix in one cluster.
- Access your cluster — kubeconfigs in CI, the API, and your own identity provider.