Launching today
Defang Station
Always-on Claude Code and Codex agents with Superpowers.
19 followers
Always-on Claude Code and Codex agents with Superpowers.
19 followers
Superpowers for Claude Code and Codex: a computer of their own that doesn't go away. Your agents can react to GitHub, Linear or your app while you're out. Your site stays live at a link you can share and ships to your own cloud with Defang Deploy. Sign in with the Claude or ChatGPT plan you already pay for, and drive it from their own apps. Built on Agent Box, the MIT-licensed runtime you can self-host. $39/month flat, no token markup. First hour free, then a 7-day trial.
Hey Product Hunt!
Raph from Defang here.
We're a team of three, and we built Station for folks who've outgrown app builders and now run Claude Code, Codex or both.
Claude Code and Codex both run in the cloud now, so why a Station? Their clouds are good for handing off a task and getting a pull request back. But each new session starts fresh from your saved setup, with nothing still running from the last one, and each cloud runs only its own agent. A Station is your agents' own computer, and it doesn't go away. Claude Code and Codex work on the same files. So one can build while the other checks. And your repos, tools and signed-in accounts are still there tomorrow. Pick it back up from Anthropic's or OpenAI's apps, on your laptop or your phone (as you can tell from the video, I'm a fan of the phone-based workflow 😁).
So it can do three things:
Publish what it built. Ask your agent to publish it and you get a real link. Open it on your phone or send it to a customer. It stays up because your Station does.
Then ship it for real. When it's ready, Defang Deploy takes the same project to production on the cloud and region you pick, in your own account. Ash Tiwari runs ekai, a private AI workspace for regulated teams:
"I used Station to take full control of ekai as a solo founder. Rebuilding the user experience, smoothing out how we deploy into our own cloud, and making it quick to stand ekai up inside a customer's cloud, which is what our enterprise buyers actually ask for."
Stop being the relay! (I mean... we've all found ourselves playing meat-proxy sometimes... 🥩📞) Instead of waiting for something to break and then copy/pasting logs around, ask your agent to connect your Station to GitHub, Linear or your own app. In the video, I hook up a Lambda to ping Codex when something crashes. Those events go straight to your agent, on the Station where your code already lives. An agent can also watch the repo and issues it's working on until the work is done. Have it immediately sort new issues, or take a first pass at a PR before you look.
Answers to some questions you might be asking:
You sign in to Claude Code and Codex yourself, inside your Station, with your own accounts. The provider's own CLI saves the login there, and Defang never receives it (more at https://station.defang.io/docs/security/). We don't resell tokens, so your AI bill stays yours.
$39/month flat. The first hour is free with no card; adding a card starts a seven-day trial.
One Station per account for now, hosted on Azure in the US, Canada or the EU, depending on where you sign up.
Your Station is usually ready right away. At busy times, building a new one takes about two minutes.
It runs on Agent Box, our MIT-licensed runtime. Read the code, or self-host it: https://github.com/defangdevs/agent-box/
The video shows what you can do today. Next we're making it easier to get your own software doing the same: reacting to bugs and user feedback, with your agents doing the work.
What would you give your agent first if it had a place of its own?