Framer AI AgentsDesign and publish professional sites with AI
Speculos
Type "deploy" in Claude Code. Your app is live.
23 followers
Type "deploy" in Claude Code. Your app is live.
23 followers
Type "deploy" in Claude Code and your app is live at a shareable link ~15 seconds later - frontend and backend. Free for frontend apps, no sign-up. Teams get their own domain, link permissions, and company data connections. Your agent can deploy on its own, too.
Hey Product Hunt 👋 Aboud here, with my co-founder Peter.
Before this, I was a product lead at a tech company. On my team, anyone - designers included - could push to a branch and it auto-deployed, so sharing a prototype was trivial. Beyond my team, I started noticing non-technical folks learning Claude Code, building things that actually worked… and hitting a wall the moment they wanted to show someone. Some were literally recording Loom videos of an app running on their own laptop, because that was the only way to share it.
That felt absurd. I am now building the fix with Peter.
You install one plugin, and from then on you just type "deploy" in Claude Code. About 15 seconds later your app is live at a link you can send to anyone. No account, no keys, nothing to configure. The app builds on your machine - we only receive the finished app, never your source code. Your agent can run deploy on its own, too.
It's free - unlimited frontend apps, no sign-up: https://speculos.ai/start
If you need backends, pointing to your own domain, link permissions, company data your admin connects once, talk to us.
Peter and I will be here all day. Try it and drop the first link you publish in the comments. We'd love to see what you make.
the "your agent can deploy on its own too" part is where I'd want a safety net. vibe-coded apps are notorious for accidentally baking API keys or secrets into the frontend bundle, and if an agent can trigger deploy without a human glancing at the build first, that mistake ships to a public link in 15 seconds instead of staying local. do you scan the built output for anything that looks like a leaked secret before it goes live, or is it purely a pipe that publishes whatever gets built
@galdayan Great point! That is one of the things we do that abstracts it away for non-technical users. We do a scan on all variables and endpoint calls and either remove or replace them. This is required for our product to work, as by default we have to rewrite the endpoint calls to what is deployed from localhost. On our enterprise plan, we have a connection broker where, if an app is calling an API to make a dashboard, we route it through our platform, enforcing auth policies and removing the creds.
how does the backend hosting work under the hood, is it serverless functions or a full container running for each project?
I like that you're reducing deployment to a single intent instead of another workflow.
As AI agents become capable of building software, I think deployment stops being a DevOps task and becomes an execution primitive the agent invokes whenever it's ready.
That feels like a meaningful shift.
Gave it a try with some static website and it worked as promised in less than a minute. Really nice!
Testing it for some of my projects, very interesting niche!
@dreb great that you tried it!
Finnt
Great stuff! solving the last mile is what matters. Most apps never see light because of deployment!