---
title: Deploy a container image
description: Run an application from an existing container image and tag, with its ports, resources and health checks.
---

If you already build your images, in CI or elsewhere, deploy them as they are: Kapten pulls the image and runs it.

1. **Create the application**

    In your environment, create an application and choose **Container** as its source. Enter the image name, such as `ghcr.io/acme/api`, and its tag, such as `1.4.2`.

2. **Set the runtime command**

    Optional: override the image's entrypoint and arguments.

3. **Set resources and instances**

    Set CPU and memory requests and limits, for example a request of `250m` CPU and `256Mi` memory. See [Resources](/reference/resources) for the units. Set the minimum and maximum number of running instances.

4. **Declare ports and volumes**

    Add each port the application listens on, with its protocol, and mark one as the default. Add a persistent volume for any path whose data must survive a restart.

5. **Add health checks**

    A **readiness** check tells Kapten when an instance can receive traffic; a **liveness** check restarts an instance that stopped responding. For a web service, an HTTP check on a health endpoint such as `/healthz` is usually enough for both.

6. **Create and deploy**

    Kapten deploys the application as soon as it is created. Open it to follow the run and read its logs.

## Deploy a new version

Change the application's tag and deploy it again. To deploy several applications in a set order, use a [pipeline](/guides/build-a-pipeline).

> **Tip**
>
> Use immutable tags, such as a version or a commit SHA, rather than `latest`: each run then records exactly what it deployed.
