PROHOSTIVA
Documentation menuAI-built websites
Deploy

Deploying a website built with AI

Built your site with Lovable, Bolt, v0, Cursor, Claude Code or Codex? Get the code out of the tool and put it online with ProHostiva.

The short version

AI tools write real code: usually a React, Vue or Next.js project, sometimes a small Python or Node server. ProHostiva doesn't care which tool wrote it. It looks at the files and deploys what's actually there. So the job is simple: get the code out of the tool, and give it to ProHostiva as a ZIP or through GitHub.

There's no plugin or direct integration with these tools, and you don't need one. If you can download your project or push it to GitHub, you can deploy it.

1. Get the code out of the tool

Every tool offers at least one of these: a download, or a connection to GitHub.

Built withHow to get the code
LovableConnect the project to GitHub from inside Lovable, then deploy that repository. Step by step: Lovable on your own domain.
BoltDownload the project, or push it to GitHub. Step by step: Bolt on your own domain.
v0Download the code, or send it to a GitHub repository. Step by step: v0 on your own domain.
Cursor, Windsurf, Claude Code, Codex, GitHub CopilotThe code is already on your computer (or in a GitHub repository). Zip the project folder, or push it to GitHub. Step by step: Cursor, Windsurf, Claude Code, Codex, GitHub Copilot.
ChatGPT or Claude (chat)Save the files into one folder, then zip it.

Which one should you use? GitHub is best if you'll keep changing the site: each change is one click to redeploy. A ZIP is fine for a site you finish once. Both are covered step by step: Deploying from GitHub and Deploying from a ZIP.

2. Check it before you deploy (free)

AI-built projects often work in the tool's preview but have a few gaps for real production: a missing setting, a key pasted into the code, a template page title. The free production readiness check builds your project the way a deployment would and tells you what to fix, in plain language. It takes a few minutes and needs no account.

3. Deploy it

  1. Create a website

    In your dashboard, click Host a website, name it and pick its address.

  2. Upload the ZIP or choose the repository

    On the website's page, open Developer settings → Deployments. Pick ZIP or GitHub as the source.

  3. Review what we detected, then Publish

    We show the framework and how it will be built and run. For most AI-built projects you don't need to change anything. Click Publish and follow the deployment log.

4. Add your settings (API keys, database URLs)

If your project uses a service such as Supabase, Firebase, Stripe or an AI API, it needs its keys and addresses. In the tool they were set for you; on ProHostiva you add them once, as environment variables: open Developer settings → Environment variables, enter the name (for example VITE_SUPABASE_URL) and its value, and deploy again.

  • Look for a file called .env or .env.example in your project: it lists the names your code expects.
  • Tick Secret for anything private. Secret values are stored encrypted and never shown again.
  • Don't put a .env file with real keys in your ZIP or repository. Anyone who sees the files can read them.
For developers

Variables starting with VITE_, NEXT_PUBLIC_, NUXT_PUBLIC_ or PUBLIC_ are passed to the build, because frameworks bake them into the JavaScript sent to the browser. That also means they are public: never give a secret key one of these names. All other variables are only given to your running app. PORT is set by ProHostiva.

Common problems with AI-built projects

  • The project is in a subfolder. ProHostiva reads the top level of your ZIP or repository. If your app lives in, say, my-app/frontend/, zip that folder, or move the app to the root of the repository.
  • The page is blank after deploying. Usually a missing VITE_ variable (see above). Add it and deploy again.
  • It works in the tool but the build fails. Open the deployment log, fix the first error it shows (often a missing package), and deploy again. See Troubleshooting.
  • Newer Lovable projects built on TanStack Start are recognised, but aren't yet in our end-to-end tested list (see Supported frameworks). If yours doesn't deploy, email us and we'll help.