---
title: Clusters
description: Kapten deploys into Kubernetes clusters in your own cloud account; you connect each one with its kubeconfig and Kapten installs an agent on it.
---

A **cluster** is a Kubernetes cluster that Kapten deploys to. Kapten does not host your workloads: your applications, jobs and databases run on clusters in your own cloud account, so your data stays in your infrastructure and your cloud bill stays yours.

## Connecting a cluster

You connect a cluster with:

- a **name**,
- its **cloud provider**: `aws`, `gcp`, or `custom` for any other Kubernetes cluster,
- its **kubeconfig**, which Kapten uses to install and update its agent.

Kapten then installs its **agent** on the cluster, with the components it relies on: an ingress gateway, cert-manager for TLS certificates, and collectors for logs and metrics. They live in namespaces whose names start with `kapten`. The agent carries out deployments for the Engine and reports their progress and logs back. See [Connect a cluster](/guides/connect-a-cluster).

## Cluster state

| State | Meaning |
| --- | --- |
| Waiting for agent | Kapten has not heard from the cluster yet. |
| Bootstrapping | Kapten is installing its agent. It takes a few minutes. |
| Connected | The agent is installed and the cluster accepts deployments. |
| Updating | Kapten is updating its agent on the cluster. |
| Bootstrap failed | Kapten could not install its agent: check the kubeconfig's credentials and that the cluster's API server is reachable, then sync. |
| Update failed | The last update failed. Sync to try again. |

**Sync** reinstalls or updates the agent. Replacing a cluster's kubeconfig also starts a sync.

## Environments on a cluster

Each [environment](/concepts/environments) runs on exactly one cluster, and a cluster can host the environments of many projects. A common setup is one cluster for production and one shared by development and staging.
