A pre-launch check for apps built with AI
We find it. Your agent fixes it.
You built this with an AI agent, and it works. Here is what that setup usually skips.
Paste one prompt (instructions you paste into your AI agent), re-scan, done.
Technical information, not legal advice.
Paste your URL. Every finding in plain words, and the prompt to fix it.
What we check
Six things, all read from what your site already shows in public. Nothing is tested against your database or your login.
1. Exposed keys
Your live JavaScript, checked for keys that were meant to stay on a server: Stripe, OpenAI, Anthropic, AWS, and Supabase's service_role key, among others.
2. Public files
Whether .env, .git, a backup, or a source map can be read at your domain. Any one of them hands over your whole setup.
3. Security headers and cookies
The headers your server sends, and whether your cookies are marked secure. Most AI-generated setups skip both.
4. Privacy and legal pages
Whether your site has a privacy policy, a cookie notice, terms, and a public accessibility statement, and which one is missing.
5. Accessibility
Every element that fails WCAG (the technical rules for an accessible website), the same check we run on any site.
6. Speed, SEO, and expiry
How fast your page loads, whether Google can actually index it, and when your domain and your SSL certificate expire.
One prompt to fix what it finds
The report ends with a single prompt, written for your AI tool, not for you.
It lists what failed, the corrected code where we have it, and the file or line the fix goes in.
Paste it into Lovable, Bolt, v0, Cursor, Claude Code, or Replit, and let the same tool that built the app apply the fix.
We only read your site. Your AI tool changes the code; we check it again afterward.
Check it from where you already build
Same report, three doors.
- 1. On the web
Paste your URL. No account, about two minutes, and the report with the prompt.
- 2. In your terminal
npx @rampa-solutions/cli check your-app.com. It exits with an error code, the same way a failed test would, so a critical finding can block your deploy.
- 3. In Claude Code or Cursor
rampa mcp adds Rampa as a tool inside your AI editor. Ask it to check the site before you ship, without leaving the editor.
Questions people ask
- Do I need to know what WCAG or ADA mean to use this?
- No. WCAG (the technical rules for an accessible site) and ADA (the U.S. law that requires one) are explained in plain words in the report, next to each failure.
- Does this replace the security checks inside Lovable, Bolt, or Replit?
- No, and it does not need to. Those tools check your code before it ships; we check what a stranger sees after it ships, from the outside, the same way a visitor would.
- If I fix everything you flag, can you guarantee I will not get sued?
- No. Nobody can guarantee that. What we can tell you is whether your code fails the rule a lawsuit cites, and we re-test every fix against that same rule.
- Do you change my code?
- No. We only read what your site already shows to any visitor. The prompt tells your AI tool what to change; you decide when to apply it.
- Is this legal advice?
- No. It is a technical check, and the report says so. Talk to a lawyer for anything specific to your business.