Open alpha · what that means

The assembled Kubernetes platform.

Create an account Talk to the platform team

Create a managed cluster with the Kupe CLI.

~/hello-kupe — kupe cli
kupe auth login --tenant hello-kupe --set-default ✓ Logged in to tenant Hello Kupe as hello@kupe.cloud kupe cluster create hello-kupe --cpu-limit 2 --memory-limit 4Gi --storage-limit 10Gi ✓ cluster hello-kupe ready kupe cluster kubeconfig hello-kupe --merge --exec ✓ Kubeconfig merged. Current context: hello-kupe Name: hello-kupe Type: shared Phase: Running Endpoint: https://hello-kupe.clusters.kupe.cloud CPU: 2 Memory: 4Gi Storage: 10Gi kubectl apply -f hello-kupe-app.yaml Application configuration applied GitOps syncs the application release and applies its database migrations. kubectl rollout status deployment/hello-kupe -n hello-kupe deployment "hello-kupe" successfully rolled out
You're driving — click windows, dock, or hotspots.
Console, GitOps and Monitoring are captured from the Hello Kupe deployment. The game is live. CLI steps are illustrated; timings are compressed. Hello Kupe deployment walkthrough. CLI steps are illustrated; timings are compressed. Create your Kupe account →

[01] Cluster running in 1m 31s

Running in about 90 seconds. Platform included.

Create a cluster from the console, the CLI or Terraform and it reports Running, with dashboards, logs, alerts and GitOps wired in, before most managed control planes have finished provisioning.

Kupe: 1m 31s. Azure AKS: 6m 18s. Google GKE: 5m 00s. AWS EKS: 11m 26s.

7.5x faster to a running cluster than AWS EKS*

With a node already available. Starting a new node first adds about two minutes.

* Kupe: median of 15 production runs of our end-to-end suite between 10 September and 16 October 2026, from the create request to the cluster reporting Running (2 vCPU, 4 GiB, single node) with a node already available. When the node pool had to start a new node first (13 runs) the median was 4m 48s, about two minutes of which was the node booting. AWS and Azure: wall-clock terraform apply for a cluster with a three-node pool, measured by Mattias Fjellström in July 2025 (eu-north-1 11m 26s, Sweden Central 6m 18s). Google: published estimate of about five minutes, not measured. Your times will vary by region and configuration.

[02] Why Kupe

Simple to start. Still Kubernetes.

Teams usually choose between an app hosting platform and running Kubernetes themselves. Kupe sits between the two: all of Kubernetes, with the platform work done for you.

An app hosting platform

Push an app, get a URL.

Quick on day one. When you need deeper infrastructure control, your options depend on the platform.

Kupe Cloud

A cluster that arrives with the platform running.

Create a cluster and deploy. Dashboards, logs, alerts and guardrails are already switched on, and we keep the platform patched. When you need more, it is still Kubernetes, so Helm charts and operators just work.

Kubernetes you assemble

A control plane and an empty cluster.

GitOps, dashboards, logs, alerting, policies, secrets and ingress are yours to install, wire together and keep patched before the first service ships.

Compare Kupe with a VPS, developer platforms and bare Kubernetes →

[03] Our platform

Managed clusters. Full control.

Three layers: what each cluster gets, what your tenant shares, and what Kupe keeps running.

Per cluster

Your own API server, full admin access, standard tooling.

  • API server
  • Full admin access
  • HTTP routes
  • Default alerts
  • Custom metrics and alerts

Per tenant

Shared by every cluster in the tenant.

  • Monitoring workspace
  • Central metrics
  • Central logs
  • GitOps project
  • Custom dashboards
Tenant boundary

Run by Kupe

Installed, patched and upgraded for you.

  • GitOps
  • Monitoring
  • Logging
  • Alerting
  • Secrets management
  • Security policies
  • Runtime protection
  • Vulnerability scanning

Kupe manages the Kubernetes control plane.

See the core concepts

[04] Security

Security paths built into the platform.

Admission policies, image scanning, network isolation, secrets, tenancy, and runtime signals should be visible without turning first use into a security project.

Cluster policies

Admission-time guardrails for risky workload patterns such as privileged pods and host mounts.

Image scanning

Vulnerabilities, misconfigurations, and exposed secrets should be visible where teams already work.

Network isolation

Tenant boundaries, cluster access, and platform internals need to stay clear from day one.

Runtime protection

Runtime signals should help teams spot suspicious activity without asking them to assemble their own detection stack first.

[05] Platform journey

From cluster to app.

Kupe is designed around the path teams actually need: create a cluster, deploy a workload, inspect the signals, and keep enough control to use the Kubernetes tools they already know.

01

Create your cluster

Sign up, name your tenant, and launch a managed cluster from the console, the CLI, or Terraform.

02

Deploy the way you want

Use kubectl, Helm, or GitOps without learning a proprietary deployment model.

03

Observe and extend

Check logs, metrics, dashboards, alerts, and routing, then extend the platform with the tools your team already uses.

  1. Closed alpha Built and tested with invited teams.
  2. Open alpha We're here Free to use, no card to add. Try it and tell us what you think.
  3. General availability Shaped by what we learn in alpha.

Create a Kupe Cloud account.

Kupe Cloud is in open alpha and free to use, with no card to add. There is no SLA yet, please read the Alpha Service Notice first.

Create an account