Infrastructure, VMs, and GPU billing

Run your own cloud,
on your own servers

Manage bare metal, VMs, GPUs, databases, and app deploys from one place. Built on Kubernetes.

$ curl -fsSL https://get.runos.com/cli.sh | bash
v1.0 · generally available 20+ managed services 6 ways to deploy runs behind NAT · zero inbound ports standard kubernetes underneath

What do you want to run?

Explore all use cases →
the same cluster ·
nodes ·
cpu / 8 gb each · per month
hetzner
$74
with RunOS
linode
$360
4.9× more
digitalocean
$480
6.5× more
google cloud
$1,020
13.8× more
aws
$1,180
15.9× more
Same cluster. 94% cheaper. pricing approximate ·

From app deploys to your GPU fleet

Explore the workflows in the terminal. The app tab also includes a recorded demo.

runos · app deploy Recorded demo
local files → in-cluster build → rollout Read the guide →

Everything in the box

Databases, queues, storage, AI, deploys, and the Kubernetes that runs them. Without the Kubernetes.

Databases

PostgreSQL, MySQL, ClickHouse. Vector, graph, and document-DB extensions on Postgres. HA tiers, read replicas, and configurable backups.

Infrastructure

Valkey, Kafka, RabbitMQ, MinIO (S3-compatible object store), Harbor registry, and more. 20+ managed services and growing.

AI stack

vLLM, Ollama, LiteLLM, and Langfuse on your own GPUs. Nvidia drivers and MIG slicing handled for you.

Vibe Deploy

Tell RunOS what to deploy using natural language. AI figures out the rest. No Git or Docker needed.

Deploy from your laptop

Ship straight from local files: no local git, no local Docker, no registry. The cluster builds your code.

CI, your way

GitHub Actions or GitLab CI with generated workflows. Or bring your own GitHub app and self-hosted runners.

Auto cloud provisioning

Add a Hetzner, DigitalOcean, Scaleway, Lambda, or Hyperstack key and RunOS provisions everything.

Any machine, even behind NAT

Join any Ubuntu server with a single curl command. Agents dial out, so zero inbound ports are needed.

Global clusters

Provision anywhere in the world. Mix providers, regions, cloud and bare metal in one cluster.

Zero K8s knowledge

Manage everything through the console, CLI, or AI/MCP. Best practice manifests crafted automatically.

Power without the headache

Kubernetes HA scheduling, self-healing, and load balancing built in. RunOS handles the complexity.

Observability included

Prometheus alerts, Grafana dashboards, and log search out of the box. Dashboards are VPN-gated by default.

Your cluster, your rules

Start with a RunOS Cloud server or bring your own. You choose where your apps and data run.

Standard Kubernetes underneath

RunOS writes plain manifests to your cluster. No forked distro, no proprietary runtime.

Your kubeconfig

Full kubectl access stays yours. Self-deploy anything alongside what RunOS manages.

Disconnect any time

On infrastructure you bring, your cluster keeps running if you disconnect RunOS.

001 How to bill for GPU usageWhat to meter, how to price on-demand, spot, and inference, and what a GPU invoice should show. → 002 It's been a while: what we've shipped since 1.0New RBAC with hardware-key approvals, BMC management for bare metal, VMs on KubeVirt, GPU inference pools, and the pieces for running your own small cloud. → 003 RunOS 1.0 is generally availableEvery component graduated to v1.x. Beta users upgraded in place, nothing broke. → 004 Deploy from your laptop with nothing installedNo local Docker, no git remote, no registry. The cluster builds your code in an ephemeral BuildKit pod and serves it over HTTPS. → 005 Zero inbound ports: reaching servers behind NATAgents dial out over mTLS, so your servers need no public IP and no port forwarding. Your basement box counts. →