Nov 28, Paris: AI Health Hack #2 with Alan, Qovery and AWS. Applications close Nov 21

Try Qovery on Your Laptop: Ask Your Coding Agent, Get a Running App

The Qovery agent skills now create a free local Kubernetes cluster in Docker when your organization has none, then deploy your project to it. One prompt in Claude Code, Cursor or OpenCode, no cloud account, no credit card. Add a database, clone a staging copy with its own URL, and share it with ngrok or Cloudflare Tunnel.

Romaric Philogene
CEO & Co-founder
NOV 9, 2026 · 8 MIN
Try Qovery on Your Laptop: Ask Your Coding Agent, Get a Running App

Key Points:

  • Your coding agent can now try Qovery for you on your laptop. When your Qovery organization has no cluster, the Qovery agent skills offer to create one locally with qovery demo up, then build and deploy your project to it.
  • One prompt does it: "Deploy https://github.com/traefik/whoami with Qovery on a local cluster on this machine." You sign in, pick your organization and, on macOS, type your password once.
  • It goes past a single app. On the same laptop we added a PostgreSQL database in about 70 seconds and cloned the whole environment into a staging copy, with its own URL and its own database, in 67 seconds.
  • Share what you deployed with ngrok or Cloudflare Tunnel, so a teammate or a webhook can reach it.
  • No cloud account and no credit card. You need Docker, a free Qovery account and a coding agent that supports Agent Skills. When you are ready for production, the same project deploys to a cluster in your own AWS, GCP, Azure or Scaleway account.

Qovery · Infrastructure control plane
Build with Claude Code, Deploy with Qovery
Learn more

Qovery in two paragraphs

Qovery is an Agentic DevOps platform. It deploys and runs your applications on Kubernetes clusters that you own: in your AWS, GCP, Azure or Scaleway account, or a cluster you already run. You describe what to run (a Git repository, a container image, a database) and Qovery builds it, deploys it, gives it a URL with TLS, and keeps every environment in one place. You don't write Kubernetes manifests or Helm charts, and you don't need a DevOps engineer to get started.

You can drive it from the web Console, the CLI, Terraform, the API, or a coding agent such as Claude Code, Cursor or OpenCode with the Qovery agent skills installed.

Until now, trying it meant connecting a cloud account or a Kubernetes cluster before you could deploy anything. That is a lot to ask of someone who only wants to see whether Qovery fits: cloud credentials or a running cluster, and about 30 minutes for a new cluster to come up, before the first deployment.

We already had a way around it: qovery demo up, a CLI command that creates a k3s cluster in Docker on your machine and connects it to your Qovery organization. But you had to know it existed, install its dependencies and run a handful of commands in the right order. Now your coding agent does that for you.

What changed

The Qovery agent skills (qovery-deploy, qovery-onboard and qovery-signup) check your organization for clusters before they deploy. If there is none, and you did not say the deployment is for production, the agent offers two paths:

  1. Try it now on this machine: a local cluster in Docker, free, for testing only. About 5 to 10 minutes.
  2. Use your cloud or your own Kubernetes cluster: AWS, GCP, Azure, Scaleway, or a cluster you already run.

If you pick the first one, the agent checks Docker, k3d, helm, kubectl and your Qovery CLI version, tells you what is missing in one message, and runs qovery demo up. When the cluster is ready, it creates a project and an environment, adds your repository, builds it on the cluster and gives you the URL.

A few words you will see:

  • Cluster: the Kubernetes cluster your apps run on. Here, a k3s cluster inside Docker on your laptop.
  • Project: a group of environments for one product.
  • Environment: a full copy of your stack (apps, databases, variables), for example dev, staging or production. Each one gets its own URLs.
  • Service: one thing that runs in an environment: an application, a container, a database, a job.

How to try it

You need Docker running with at least 4 CPUs and 8 GB of memory (macOS, Linux, or Windows with WSL), a free Qovery account from console.qovery.com, and a coding agent that supports Agent Skills.

Install the skills once:

BASH
curl -fsSL https://skill.qovery.com/install.sh | bash

Then open Claude Code, Cursor, OpenCode, Gemini CLI or another agent that supports Agent Skills, and paste:

TEXT
Deploy https://github.com/traefik/whoami with Qovery on a local cluster on this machine.

traefik/whoami is a small public web server, a good first run. Swap the URL for any repository with a Dockerfile, or open the agent in your own project and say "Deploy my application with Qovery". If your project has no Dockerfile, the agent writes one.

The agent hands a few steps to you, because they open a browser or need your password:

StepWhy you run it
qovery authOpens a browser to sign in to your Qovery account
qovery context setAn interactive picker for your organization
macOS: one sudo ifconfig commandAdds a local network address so your browser reaches the apps, and needs your password
WSL: qovery demo upThe setup needs sudo and a Windows administrator prompt

In Claude Code, type ! before a command to run it in the session, for example ! qovery auth.

At the end the agent gives you an HTTPS URL for your app. It has a valid certificate and opens in your browser like any website. You also see the cluster, the environment and the deployment logs in the Console.

What we measured

We ran it from scratch with a fresh Claude Code session that had nothing but the Qovery skills installed, on an Apple M3 Ultra, Qovery CLI 1.181.1, on 9 and 10 November 2026:

StepTime
qovery demo up creates the local cluster4 min 43 s
The agent deploys a public container imageabout 6 min for the whole agent run
The agent builds traefik/whoami from GitHub and deploys it2 min 40 s for the whole agent run, about 1 min 30 s from build start to running
Add a PostgreSQL 17 database to the environmentabout 70 s from deploy to running
Clone the environment, app and database, into staging67 s from deploy to running, HTTP 200 on its own URL

The build runs on the local cluster itself. Your times will depend on your machine, your network and the project.

The runs were not perfect, and we fixed what we saw. In the first runs the agent passed a file path where the CLI expects stdin, tried the latest image tag that the API rejects, and read runtime logs when it wanted build logs. It recovered each time, and the skills now say all three things up front.

Try Qovery on your laptop today.
Install the Qovery skills, then ask your coding agent to deploy a GitHub repository. Qovery Free: 3 users, AI agents don't count, 1 cluster, 1,000 deployment minutes a month, no credit card.

More than one app: databases and testing environments

A real project is rarely one container. On the local cluster you get the same building blocks as on a production cluster.

What you can deploy:

  • Applications from Git: Qovery builds the image from your Dockerfile on the cluster. GitHub, GitLab and Bitbucket.
  • Container images from Docker Hub, GHCR, ECR or any other registry.
  • Databases: PostgreSQL, MySQL, MongoDB and Redis, running as containers on the cluster. Qovery creates the credentials and injects the connection details into your apps as environment variables.
  • Jobs and cron jobs: database migrations, seed scripts, nightly tasks.
  • Environment variables and secrets, set per service, per environment or per project.

Ask your agent for them in plain words:

TEXT
Add a PostgreSQL database to this environment and give the app its connection URL as DATABASE_URL.

On our run, the PostgreSQL 17 database was running about 70 seconds after the deploy started.

Testing environments with their own URLs. Once dev works, clone it. A clone copies every service and variable into a new environment, and each service gets its own URL, so you can test a branch or a migration without touching the original:

TEXT
Clone this environment into a new environment called staging and deploy it.

Our clone, with the app and its own fresh PostgreSQL database, was running 67 seconds after the deploy started, at a new URL. The data is not copied: the clone starts with an empty database, which is what you want for a test run.

Delete it when you are done, or ask the agent to. With Git repositories connected to Qovery, you can also turn on preview environments, which create one environment per pull request and delete it on merge.

Share your local app with ngrok or Cloudflare Tunnel

The URLs of the local cluster point to an address on your own machine, so they only open on your laptop. To show your app to a teammate, test it from your phone, or receive a webhook from Stripe or GitHub, put a tunnel in front of it. The tunnel runs on your laptop and gives you a public URL that forwards to the app.

Take the app URL from the agent or the Console, for example https://zcd88b9a5-z08cdc690-gtw.z40c5ed14.qovery-demo2.com. Each app is routed by its hostname, so tell the tunnel to send that hostname to the cluster:

Cloudflare Tunnel (free, no account needed for a temporary URL):

BASH
cloudflared tunnel --url https://<app-host> --http-host-header <app-host>

It prints a https://<random-words>.trycloudflare.com address.

ngrok (free account and auth token needed):

BASH
ngrok http https://<app-host> --host-header=rewrite

It prints a https://<random>.ngrok-free.app address.

Replace <app-host> with the hostname part of your app URL, without https://. Anyone with the tunnel URL can reach your app while the command runs, so stop it with Ctrl+C when you are done. Or ask your agent: "Expose the whoami app with a Cloudflare tunnel".

Why start on a laptop

A local cluster answers the first question a developer has, "does my app run on this?", in about ten minutes and without asking anyone for cloud access. It is the same Qovery: the cluster shows up in the Console next to any other, with its environments, deployment logs and variables. What you set up there, you keep when you move to production.

When you are ready, create a cluster in your own cloud account, or connect one you already run, and deploy the same project to it. Your agent can do that too: "Deploy this to production on AWS with Qovery".

What it is not

The local cluster runs only on your machine and is not for production. It stops when Docker stops, and apps you deploy there are reachable from your machine only, unless you open a tunnel.

On the Free plan, the local cluster counts as your one included cluster. To connect a cloud cluster later, remove the local one with qovery demo destroy --delete-qovery-config, or move to a paid plan to keep both.

The local cluster needs Docker with at least 4 CPUs and 8 GB of memory, on macOS, Linux or Windows with WSL. Each extra environment uses some of that memory, so delete the ones you no longer need.

Get started

The step-by-step guide is in the docs: Try Qovery on your laptop with a coding agent. If you prefer to run the commands yourself, without an agent, see Test & Learn Locally.

Frequently asked questions
Do I need a cloud account to try Qovery?

No. The local cluster runs in Docker on your machine. You need a free Qovery account, Docker, k3d, helm and kubectl.

Which coding agents work?

Any agent that supports Agent Skills: Claude Code, Cursor, OpenCode, Gemini CLI and others. Install the skills with curl -fsSL https://skill.qovery.com/install.sh | bash.

Can I run a database on the local cluster?

Yes. PostgreSQL, MySQL, MongoDB and Redis run as containers on the cluster, and Qovery gives your apps the connection details as environment variables. A PostgreSQL 17 database took about 70 seconds to start on our test machine.

Can I create several environments, like dev and staging?

Yes. Clone an environment and Qovery copies its services and variables into a new one, with its own URLs and its own empty databases. Our clone took 67 seconds to deploy.

How do I share an app running on my laptop?

Use a tunnel such as Cloudflare Tunnel or ngrok, pointed at the app's HTTPS URL with the app's hostname as the host header. The commands are in the section above.

Is it free?

Yes. The Qovery Free plan has no credit card and no time limit: 3 users, AI agents don't count, 1 cluster and 1,000 deployment minutes a month. The local cluster uses that one cluster.

Can I use the local cluster in production?

No. It runs on your laptop and is meant for trying Qovery and testing. For production, create a cluster in your own AWS, GCP, Azure or Scaleway account, or connect a Kubernetes cluster you already run.

How do I remove it?

Ask your agent, or run qovery demo destroy. Add --delete-qovery-config to also remove the cluster and its environments from your Qovery organization.

Romaric Philogene
About the author
Romaric Philogene

Romaric founded Qovery to make Kubernetes accessible to every engineering team. He writes about platform strategy, developer experience, and the future of cloud infrastructure.

Next step

Try Qovery on your laptop today.

Install the Qovery skills, then ask your coding agent to deploy a GitHub repository. Qovery Free: 3 users, AI agents don't count, 1 cluster, 1,000 deployment minutes a month, no credit card.