Exporting code from an AI tool is straightforward, but the export format determines where that code can actually run. Chat assistants produce isolated files or code blocks. Visual app builders produce complete front-end projects that rely on external databases or temporary asset URLs. Code editors keep files on your local drive from the start.
Before moving files to your Flashcloud hosting account, you need the right bundle type: static production files for the document root, or a complete project directory for the Node.js application runner. Here is how to pull your code from eight popular tools, what those exports leave behind, and where to put them on your server.
Chat assistants: ChatGPT, Claude, Gemini, and Perplexity
Conversational assistants generate code on demand. Because they do not compile projects inside a local filesystem, you must copy or download their outputs manually, or connect them to your account.
- ChatGPT: If using Canvas, click the menu in the upper corner of the code pane and select download, or copy the block directly. ChatGPT produces individual script files, snippets, or single-file HTML/CSS/JS bundles.
- Claude: When Claude creates an Artifact, click the download icon in the bottom-right corner of the Artifact window to save the file or ZIP bundle. Alternatively, if you connected Claude using our connector at
https://mcp.flashcloud.com/mcp, you can instruct it to write files directly to your account. See managing connected AI agents in the portal for permission controls. - Gemini: Click the copy icon on the code snippet box. For larger projects generated in Gemini Canvas, export the file directly or copy the entire buffer.
- Perplexity: Copy the code block from the answer pane. Perplexity produces text and snippets rather than repository downloads.
What they miss: Chat assistants give you raw source code without running build tools. If the model outputs React JSX components instead of a bundled HTML document, you cannot drop them directly into your web root. You must place them into a local framework project, compile the production assets, and upload the build output.
App generators: Lovable, Bolt, and v0
These platforms build full-stack or multi-file front-end applications inside web containers. They offer direct download buttons and GitHub sync.
- Lovable: Click the GitHub icon or project menu in the top bar. You can either push the project to a GitHub repository or download the source as a ZIP file.
- Bolt.new: Click the download icon in the top-right toolbar to download a ZIP archive containing the complete project directory, or click the GitHub button to push the repository.
- v0 by Vercel: Click the code view, then use the download button to grab a ZIP archive. Alternatively, click the CLI tab to copy the
npx shadcncommand to pull the component into an existing local project.
What they miss:
- Database connections: Lovable applications often run against Lovable Cloud (Supabase). The downloaded ZIP contains only the front-end code. Database tables, authentication rules, and API secrets remain on the hosted database. You must recreate your tables in MariaDB or point your environment variables to a dedicated database instance.
- Temporary assets: v0 frequently uses placeholder images from external domains. Check your code for external image URLs before deploying, as these links can expire or fail under production security headers. Save your images into a local directory instead.
- In-memory backends: Bolt applications sometimes simulate database calls in-browser. Moving to production requires a real backend or static mock data.
Local development: Cursor
Cursor is an editor that runs locally on your workstation. Your code already lives on your hard drive inside your project folder.
To move Cursor projects to Flashcloud, initialize a Git repository locally, push it to a remote host like GitHub, and use cPanel Git Version Control to pull the files. If you use the Flashcloud AI connector inside Cursor, you can deploy updates directly from your editor prompt.
Matching the export to your hosting path
Once you extract your code, deploy it based on how the application was built:
1. Static HTML, CSS, and client-side JavaScript
If your export consists of an index.html file with companion CSS and JavaScript files, upload them directly to your document root. For your primary domain, this is public_html. For a subdomain, the document root is ~/<subdomain>, located alongside public_html.
Use File Manager, FTP, or Git Version Control to place the files. If your project was built using Vite or React, run your build script locally first:
npm run build
Upload the contents of the generated dist/ folder into the document root. Do not upload the whole project root. If your client-side app uses browser routing, append the following rewrite rules to the existing .htaccess file in the document root so subpages resolve correctly:
RewriteEngine On
RewriteCond %{REQUEST_FILENAME} !-f
RewriteCond %{REQUEST_FILENAME} !-d
RewriteRule ^ index.html [L]
Never delete the original .htaccess file created by cPanel. Always append your changes to the bottom of the file.
2. Full-stack Node.js applications
If your project relies on an Express server or backend API routes, it must run under the Node runner. Follow our guide to Node.js apps in cPanel to set up your environment.
Upload the application files, including package.json, to a private folder outside public_html. Open Setup Node.js App under Software in cPanel, select Node 22 or 24, define your startup file (such as app.js or server.js), and run the dependency installation. Ensure your server script listens on process.env.PORT rather than a hardcoded port number.
When to contact support
If your exported code is on the server but throwing 500 errors, check stderr.log in the application directory first. If a client assistant connected to your account fails to write files or reports permission blocks, review your agent settings in the portal or submit a ticket under Support. You can also ask your assistant to open a ticket on your behalf, as described in opening a support ticket from your AI assistant. Our support team consists of real hosting technicians who can inspect server logs and file permissions directly.