Common questions — Defang Station

What a Station is, what it costs, where it runs, and what it won’t do for you. If your question isn’t here, email us and it will be.

The basics

What is a Station?

An always-on cloud machine where your coding agents work. Sign in to Claude Code, Codex or both with your own accounts, then drive them from Anthropic’s and OpenAI’s own desktop and mobile apps. Your agents keep running after you disconnect, and your files, tools and signed-in agents are still there when you come back. Because your Station lives on the web, your agents can publish what they build at a link anyone can open, and GitHub can reach them with new issues and pull requests.

Do I need my own Claude or OpenAI account?

Yes, and that’s the point. Station doesn’t resell AI usage or mark up tokens. The agents on your Station are the official Claude Code and Codex, installed from their own published packages, and you sign in with the account you already use. Your agent bill stays between you and your AI provider.

Which agents are supported?

Claude Code and OpenAI Codex today, each signed in with your own account. The runtime is built to take more, so more will follow. Email support@defang.io with the one you need.

What happens when I close my laptop?

Nothing bad. Your agents keep running on your Station, and their sessions survive the disconnect. Pick the work back up from Anthropic’s or OpenAI’s app on your desktop or phone. Once your agent connects your Station, GitHub or your own app can even start the next job while your laptop is shut.

Can my agent put what it builds on the web?

Yes. Every Station has its own web address, and your agent can put a site or app on a subdomain of it right away. There’s nothing to configure, and HTTPS is handled for you. Nothing is published until you ask, and once it is, anyone with the link can open it, so keep internal tools off it. When you want a nicer name, point a domain you own at your Station.

Can my agents react to GitHub, or my own app, instead of waiting for me?

Yes, once you set it up. Your Station can receive events, but it doesn’t start out connected to your repositories. Sign in to GitHub on your Station, then tell your agent what to watch and what to do, for example: “Watch new issues on acme/website and open a draft pull request for each one.” The agent sets up the webhook. If your GitHub account can’t manage that repository’s webhooks, it walks you through adding it in the repository’s settings. From then on, matching events arrive on their own: an agent that’s already working picks one up, or a new session starts for it. Every delivery has to be signed with your Station’s own secret, so nothing else can trigger your agents. Linear works the same way. Your own app can wake an agent too, as long as it signs what it sends, so ask your agent to add that to your app. The security page explains what’s checked.

Do I have to work in a browser?

No. You work from the apps you already have. Once you’ve signed in on your Station, Anthropic’s own desktop and mobile apps drive the Claude session running there. For Codex, pair the Codex app on your computer or the ChatGPT app on your phone from your Station’s page.

How long does it take to get started?

Signing up starts building your Station straight away. We call it terraforming, and it usually takes about two minutes. You connect your agent and GitHub accounts while it builds.

Price and trial

What does $39/month get me?

One always-on Station, with Claude Code and Codex ready to sign in, a place to publish what they build, GitHub events that can wake them once you’ve set that up, and your work kept between sessions. There’s one plan, at a flat $39/month. The only other bill is your AI subscription, which you already have.

How does the trial work?

Your first hour is free, with no card. After that, adding a card starts a 7-day free trial and nothing is charged until it ends. The trial is once per customer, so if you have had one before, the $39 starts straight away.

Can I cancel?

Yes. Open Manage billing from your Station and cancel there. It takes effect at the end of the billing period, and your Station keeps running until then. Push your work to GitHub before that and nothing is left behind.

What are the specs?

A dedicated ARM machine sized for coding-agent work: a repository, a build, a test run and a dev server alongside your agent. ARM is the same architecture as an Apple silicon laptop, and it matters in one case: a Docker image built only for x86 will not run here. We don’t sell by vCPU. If your workload needs more, email support@defang.io and tell us what you are running.

A VPS is $10 a month. Why is Station $39?

A VPS is cheaper, and running Agent Box, our open-source runtime, on one yourself is a good setup. Station sets it all up for you: the server, a network locked down to HTTPS, an address and a certificate, Claude Code and Codex installed, and publishing and webhook routing in place. It runs on our cloud account, so there’s no server bill of your own. That’s what the $39/month pays for.

Security and control

Is my code safe? Do you see my API keys?

Sign-in happens inside your Station, through each provider’s own flow. Claude Code or Codex saves your login on your Station, the same way it would on your laptop. Defang never receives it. Every Station belongs to one account, and that is checked on every request against your signed-in session. There is one exception: a value you type into your Station’s environment settings yourself passes through Defang on its way there. It isn’t logged or kept. The security page sets out the promise and the exception together.

Where does my Station run?

On Defang’s cloud, hosted on Azure, in the United States, Canada or the European Union, depending on where you sign up from. Your Station’s page shows which, with its flag. If that region has no room when your Station is built, it goes to the next nearest one that does, and the page shows that one instead. A Station stays where it was built. If you need it somewhere else, you can self-host Agent Box, or email support@defang.io and tell us where.

Can the agent break things?

Inside your Station the agent runs with full permissions, because that’s what makes it useful. Anything it breaks stays on that one Station, away from your laptop and the systems your customers use. Station doesn’t back up your files, so keep your work pushed to Git, and review what it builds before you deploy it.

Can I run it on my own infrastructure?

The runtime, yes. Station is built on Agent Box, an MIT-licensed open-source project you can self-host on your own cloud account or your own hardware. Station is the hosted product around it, with nothing to set up. You can start with either, and nothing locks you in.

How it compares

How is this different from Claude Code on the web, or Codex in ChatGPT?

Anthropic and OpenAI each run their agent in their own cloud, and both are good for handing off a task and getting a pull request back. But both start each new session fresh from your saved setup, with nothing still running from the last one, and each cloud runs only its own agent. On a Station, you can mix and match Claude Code and Codex, and your work keeps running across sessions. Your app and its database stay up at a link anyone can open, and Defang Deploy takes the result to production in your own cloud account. You still drive your agents from Anthropic’s and OpenAI’s own apps.

How is this different from Lovable, Bolt or Replit?

App builders bundle an agent, hosting and a usage-based bill into one product. That’s a good trade until you want more control. Station is the next step. You bring your own repository and Claude Code, Codex or both, on the subscriptions you already pay for, and they work on a machine that keeps running. You keep a live link for what they build, Defang Deploy gives you a path to production in your own cloud, and the Station itself is a flat $39/month.

How does my agent’s work get to production?

Defang Deploy takes the same project to dev, staging or production, on the cloud and region you choose, in your own account. Station is where your agents build, and Deploy is how what they build ships.

Teams and support

How do I get support, or report a bug?

If you have a Station, open Send feedback from the top of any page once you’re signed in, and write to us there. If you don’t yet, email support@defang.io. Both reach the team that builds this.

Still weighing it up? See how Station compares →

Start your Station