In recent months, I felt terribly painful about deploying my work-in-progress code to a public server.

Imagine a scene where you need to go to the gym, but you don't want to waste the subscription budget from Claude or Codex, so you set up Happy or Paseo. They have done really well at remote controlling coding agents, but there is still a gap between the formative work produced by coding agents and a sharable live environment.

For the current workflow, you need to buy a server — probably from Hetzner or DigitalOcean — as the "staging server". Then you set up CI to build your app as one or several Docker images, deploy those images to the server, configure the OS, GitHub secrets, Docker Compose, DNS, Cloudflare, NGINX...

All this just to show someone a work-in-progress page.

So I built a new product named Prevu.

What is Prevu?

Prevu provides an agent-friendly experience for previewing your work-in-progress agent work.

It is not another CI platform, and it is not another PaaS dashboard asking you to write YAML before seeing anything.

The idea is simple:

Your coding agent should be able to say: “I need a preview environment.”

And Prevu should answer: “Here it is.”

With Prevu, every project can get an isolated, disposable, public preview environment. Your coding agent can start the app, expose the right port, inspect logs, and return a sharable URL.

Just a prompt.

Why this matters

Coding agents are becoming very good at editing code. They can read issues, modify files, run tests, and explain what changed.

But software is not only text. Software is behavior.

Today, agents often stop right before that moment. They produce code, but not a live artifact.

Prevu tries to close that loop.

Trying to preview a WIP branch of LangBot :

The workflow I want

Instead of:

1. Buy a server
2. Configure CI
3. Push Docker images
4. Write deployment scripts
5. Configure DNS and reverse proxy
6. Debug everything
7. Finally share a link

I want:

1. Ask your agent
2. Get a preview link

For example:

Deploy this project to Prevu and give me a preview URL.

Then the agent can create or reuse a Prevu environment, run the project, expose the port, and give you something like:

https://your-preview.prevu.page

That is the whole point.

What Prevu is for

Prevu is not trying to replace production deployment platforms like Vercel, Railway, Fly.io, or Kubernetes.

It is for the stage before production:

• before the PR is merged
• before CI is stable
• before the idea is validated
• before you want to polish deployment

It is especially useful for:

• landing pages
• dashboards
• SaaS prototypes
• internal tools
• documentation sites
• full-stack experiments
• AI-generated app previews

In other words, Prevu is for the messy middle between “the agent wrote some code” and “this is ready to ship.”

Agent-first, not dashboard-first

One design choice I care about a lot: Prevu is built to be operated by coding agents.

Of course, there is a dashboard for humans. But the core experience should work through agents like Claude Code, Codex, Cursor agents, OpenClaw, or any other coding assistant.

The best developer tools in the agent era may not be the ones with the most buttons. They may be the ones agents can use reliably on your behalf.

Why I built it

I built Prevu because coding agents made me faster at writing software, but not fast enough at showing it.

Every time I wanted to share something half-finished, I had to leave the creative flow and become an infrastructure engineer again.

That felt wrong.

If agents are going to become a real part of software development, they need a place to run their work — not just edit files.

Prevu is my attempt to build that missing layer.

Try it

Prevu is still early, but the core idea is already useful:

Give your coding agent a place to preview its work.

Try it here:

https://prevu.cloud

If you try it, I’d love to hear what breaks, what feels magical, and what your coding agent tries to do with it.