All writings
Essay · 4 min

Cursor Can Now Ship Live Web Apps: Should You Use It?

Cursor's cloud agents can now build, host and publish a web app with a database and login from a plain-English prompt. What a small team should test first.

By Peter Walters

This week, Cursor added Apps to its cloud agents. You describe a web app in plain English, and the agent builds it, provisions a Postgres database if the app needs one, sets up sign-in, and publishes it to a shareable link after a security check. Cursor's documentation says Apps is available on all paid plans with access to cloud agents.

What actually changed

Everything below comes from Cursor's Apps documentation, plus a November 6 roundup from The Neuron that listed the launch. I have not used it yet, so this is a read of the docs, not a review.

Here is what the docs say:

  • Cloud agents can build dashboards, portals and microsites, either from scratch in a new repository or from an existing GitHub repository.
  • Cursor handles hosting, the database, auth and secrets. Each app runs in its own isolated hosting account.
  • Nothing goes live until you press publish. You pick the URL, then Cursor runs security checks before giving you a live link.
  • Four checks always run: no secrets written into app files, no credential files, no .env files, and no source maps. An AI security review scans for bigger problems on top of that.
  • Visibility has three settings: people you invite, your team, or public. On Teams and Enterprise plans, public sharing is off by default until an admin turns it on.
  • Secrets are entered by you, and the docs say the agent never sees their values.
  • Scripts and bots can reach an app with a Cursor API key sent in a header.

The docs do not list pricing, usage limits or a version number, so I will not guess at costs.

Why it matters if you run a small team

Most small teams have a pile of tiny internal tools they keep meaning to build. A client portal. A booking dashboard. A page where staff log what happened at the front desk. These never get built because the gap between "a spreadsheet" and "a developer" is too wide.

Tools that close that gap are real, and I use them. I have shipped 12+ apps in a year with Lovable, Claude Code, Vercel and Supabase. So I like the direction here. The part that interests me most is the publish gate. An agent that stops and asks before anything goes public is exactly the habit I want from every AI tool.

Here is what I think is overhyped. Building an app is the easy part now. Owning it is the hard part. Someone has to decide who can see the data, fix it when it breaks in three months, and notice when a login stops working. A live link is not a finished product.

I would also be careful with public sharing. The docs say individual-plan users can share apps publicly without Cursor sign-in. That is fine for a menu or a schedule. It is a bad idea for anything with client names in it.

What I would do with it this week

Start with something boring and private.

  • Pick one internal tool you would use weekly, like a shared checklist or a simple intake form.
  • Keep it on "people you invite" or "team" visibility. Do not make it public.
  • Write down, in two sentences, what data it will hold and who may see it.
  • Let the agent build it, then open the preview and try to break it before you publish.
  • Read the list of anything the security check flags. If you do not understand a flag, ask the agent to explain it in plain language.
  • Put a date on your calendar one month out to check whether anyone actually used it.

When I built a daily brief agent for a real estate agent I work with, the build took far less time than deciding what the brief should contain. The same will be true here. Spend your first hour on the question "what is the one thing this app does," and let Cursor, Lovable or Claude Code handle the rest.

If you are not technical, ask whoever helps with your website to review the first one with you. Ten minutes of review is cheap.

Sources

Want a plain-English note like this each week? Join the letter.

If this landed

I build websites, apps, and AI tools for people who can't stop thinking about a specific thing. If that's you, let's talk.

Book a conversation