Your cluster is welded to your provider
- Managed Kubernetes from a cloud provider comes with that provider’s machines. The day you want cheaper compute, a different jurisdiction or a rack of your own, the cluster does not come with you.
- Self-managed Kubernetes on your own machines means someone on your team now owns etcd, certificates, upgrades and a control plane that falls over when a noisy workload lands on the same box.
- A migration between the two is not copying data. It is rebuilding the cluster, its access, its add-ons and everything that assumed the old one.
Create a cluster, join a machine, deploy
Pick a Kubernetes version and a region. A control plane is up quickly, reachable on its own TLS endpoint.
Run one command on any Ubuntu machine that can reach the internet outbound. It checks the host, prepares spare disks if you allow it, and joins. The token expires; it is not a standing key.
Install from a signed catalogue, or use the kubeconfig with any Kubernetes tool you already have.
From an empty cluster to a running application
Creating
Joining
A cluster
The catalogue
Deploying
Running
AI models
Standard Kubernetes, with the hard part run for you
Hosted control plane
API server, scheduler and etcd run with us, isolated per cluster, on their own endpoint. Your nodes carry only your workloads.
Bring your own node
Any Ubuntu host with outbound connectivity. No public IP, inbound firewall rule or VPN required.
Observability built-in
The cluster and its workloads are monitored automatically, with built-in alerts.
Storage on join
Optional: unused disks on a new node are formatted and handed to the cluster. Disks in use are never touched.
Application Marketplace
A marketplace of applications in a single registry: versioned, signed and verified, and deployable to any of your clusters.
Standard Kubernetes
Download the kubeconfig and use kubectl, Helm or whatever you use today. Nothing here requires our tooling.
Applications you run, solutions we answer for
Applications
We package it, sign it and keep the chart current. You deploy it and you run it. We tell you when it is unhealthy; we do not go into your cluster to fix it.
Garage
MariaDB
Microsoft SQL Server
PostgreSQL
Supabase
Solutions
We deploy it, run it and answer for it. Upgrades, incidents and tuning are ours, under an engagement.
OpenClaw
Hermes Agent
NanoClaw
OpenShell
Your machines, your data
The machines
Any provider, any jurisdiction, your own rack. Change them without rebuilding the cluster.
The data
Volumes live on your nodes. Nothing is copied to us.
The access
Standard kubeconfig, standard Kubernetes APIs.
What we hold: the control plane and the platform
This is the part we run. Kubernetes Control Plane, Automation Platform for Application Delivery, AI Gateway, LLM Inference solution, Sandboxes, AI Agents.
The ones people actually ask
Can I leave?
Yes. Your workloads use standard Kubernetes objects and charts. Moving to another Kubernetes means re-pointing them, not rewriting them.
Which operating systems can join?
Any Ubuntu server, 22.04 LTS or later. 26.04 LTS is recommended.
Do my nodes need a public IP?
No. They need outbound access to the control plane endpoint.
What does it cost?
Priced per cluster, by the size of its control plane. No limit on nodes or workloads.
See a cluster come up and a node join, on a call
- A control plane created live
- A machine of yours joined, if you bring one
- What moving an existing workload would take