How Greenlight Scan works
The short version is on the home page. This page has everything else: what we check, what we can’t see, who it’s for, every step and a full plan comparison. The four steps below are the same three steps as on the home page, with the ownership check shown as its own step.
Greenlight Scan is built for apps made with AI tools (Lovable, Bolt, Cursor, Replit, v0).
Demo with example data · See a full sample report →
- Lovable
- Cursor
- Bolt
- Replit
- v0
- Supabase
- Firebase
From red to green in four steps.
Check 1: the outside check
Paste your link. Works on any website. About 2 minutes. It looks only at what any visitor's browser can see.
Check 2: the deep check
Also free. It runs only after you show the site is yours, because it looks inside. Showing it takes one small line, and it works for any website: your AI tool, you, or whoever built your site can add it. You see your worst problem in full. Most sites get the deep check done quickly. Some hosting companies, like Vercel, slow security checks down on purpose. There it can take a day or more, but the report always finishes by itself and comes to you by e-mail.
Paste the fix
The $19 Full Scan shows every problem, with a ready-made prompt for Lovable, Cursor, Bolt, Replit, v0, Windsurf or Claude Code. Your AI does the work.
Re-scan until green
Unlimited re-scans for 30 days. All green? You get the Greenlit badge for your site.
Your app looks fine. Its insides might not.
Left: what your visitors see. Right: what anyone can pull out of their browser in two minutes. Flip the switch to see the same app after the fixes.
Example data. Leaked keys and security headers show up in the free preview; database and private-page checks run in the free deep scan after you verify you own the app; the details unlock with the Full Scan.
Pick the one that sounds like you.
You built something real. Make sure it’s not leaking.
You shipped with Lovable or Bolt and it works — but you can’t read the code to know if it’s safe. We tell you in plain words, and give you the exact sentence to paste back to your AI.
- “Can other people see my users’ emails?”
- “Did the AI put a password in my website?”
- “What if someone runs up my OpenAI bill?”
The four things that sink AI-built apps.
- Your live app, from the outside, exactly as a stranger would
- Check 1: the outside check (free). Works on any website. About 2 minutes. It looks only at what any visitor's browser can see.
- Leaked secret keys in your pages and JavaScript
- Security headers
- Cookie settings
- Insecure parts loaded on a secure page
- Outdated JavaScript libraries with known problems
- Open .env and .git files
- Your site's security certificate
- E-mail spoofing protection (SPF, DMARC)
- Whether your Supabase or Firebase setup is public
- Check 2: the deep check. Also free. It runs only after you show the site is yours, because it looks inside. On top of check 1, it looks at:
- Who can read your database and file storage (Supabase, Firebase), counting rows only
- More hidden files left open: backups, database dumps, config files and logs
- API routes that answer strangers without a login
- Open admin and debug pages, public source maps
- Login pages: a secure form, safe session cookies, protection against password guessing
- The redirect to HTTPS, open redirects and open folder listings
- Domain settings: DKIM, CAA, DNSSEC and subdomains
- Smaller hardening checks: security.txt, HSTS preload, server version, robots.txt and sitemap
You see your worst problem in full, free. The complete list with ready-made fixes is the $19 Full Scan. Why do check 2? The outside check cannot see the dangerous problems, like a database that strangers can read or files left open. It is still free.
Most sites get the deep check done quickly. Some hosting companies, like Vercel, slow security checks down on purpose. There it can take a day or more, but the report always finishes by itself and comes to you by e-mail.
- Launch Audit: 7 specialist AI reviews of the code in your GitHub repo (access control, logins, payments and webhooks, input handling, secrets, AI features, abuse), a check of every library for known vulnerabilities, a secret scan of your full git history, realistic attack paths, and a launch verdict. Every finding is double-checked by a second, independent AI
- Anything behind your login screen (we don’t log in to your app)
- Your Lovable or Bolt workspace settings
- Business-logic bugs only a human tester would notice
A green light means no known issues found — not a guarantee.
What each plan includes.
Free preview
Find out if you have a problem.
For: Anyone about to share their app link.
Two free checks. Check 1: the outside check. Works on any website. About 2 minutes. It looks only at what any visitor's browser can see. You see your score, how many problems at each severity and the title of the most serious one. Check 2: the deep check. Also free. It runs only after you show the site is yours, because it looks inside. You get your new score, the counts, and where your worst problem is and why it matters. Why do check 2? The outside check cannot see the dangerous problems, like a database that strangers can read or files left open. It is still free. Every other detail and all fixes stay locked.
Full Scan
Fix everything on your live site.
For: Founders launching, or already live with real users.
Everything from the free scan and the free deep check, shown in full. That covers which Supabase tables and storage buckets a stranger can read (row counts only, we never download data), whether your Firebase rules are open, 40+ exposed-file paths, API routes that answer without a login, CORS, admin panels, source maps and login rate limiting. Every finding comes with evidence, location, impact, a step-by-step fix and a prompt for Lovable, Cursor, Bolt, Replit, v0, Windsurf or Claude Code. Plus the passed-checks list, a PDF, 30 days of re-scans with before/after, and the Greenlit badge when you're clean.
Launch Audit
Check the code, not just the live site.
For: Apps that take payments or store personal data.
Everything in the Full Scan, plus what an outside scan can't see. Seven specialist AI reviews go through the code in your GitHub repo: access control (can users read each other's data?), logins and sessions, payments and webhooks, input handling, secrets, AI features and abuse. We check every library you use for known vulnerabilities, scan your full git history for leaked keys, and map realistic attack paths. A second, independent AI double-checks every finding. You get a one-page launch verdict, each finding with how an attacker would use it, the exact fix and how to test it, a prioritised fix plan with prompts for Cursor, Claude Code, Lovable, Bolt, Replit and v0, a detailed PDF report and a free re-test after you fix.
Free shows what’s wrong.
$19 shows how to fix it.
Scan free, verify ownership (free deep scan), then unlock from your results page.
| What you get | Free | $19 | $79 | Greenlight Watch |
|---|---|---|---|---|
| Security score + number of problems by severity | ✓ | ✓ | ✓ | ✓ |
| Quick surface checks: headers, TLS, exposed .env/.git, secrets in JS, Supabase/Firebase config, SPF/DMARC | ✓ | ✓ | ✓ | ✓ |
| Title of the most serious problem | ✓ | ✓ | ✓ | ✓ |
| Worst problem in full: where it is and why it matters | After you verify | ✓ | ✓ | ✓ |
| 40+ deep checks after ownership verification (database access, 40+ exposed-file paths, private API routes, CORS, admin panels, source maps, login rate limiting, DKIM/CAA/DNSSEC…) | Counts + worst problem | ✓ | ✓ | ✓ |
| Every problem with evidence, location, impact and step-by-step fix | – | ✓ | ✓ | ✓ |
| Copy-paste fix prompts for your AI coding tool | – | ✓ | ✓ | ✓ |
| Passed-checks list, PDF report, 30 days of re-scans with before/after, Greenlit badge | – | ✓ | ✓ | ✓ |
| 7 specialist AI reviews of your GitHub code: access control, logins, payments and webhooks, input handling, secrets, AI features, abuse | – | – | ✓ | – |
| Every library checked against known vulnerabilities (OSV) + secret scan of your full git history | – | – | ✓ | – |
| Realistic attack paths and a one-page launch verdict | – | – | ✓ | – |
| Every finding double-checked by a second, independent AI | – | – | ✓ | – |
| Prioritised fix plan with prompts for Cursor, Claude Code, Lovable, Bolt, Replit and v0, plus a detailed PDF | – | – | ✓ | – |
| Re-scans after you fix | – | 30 days, unlimited | 30 days + audit re-test | Every week, automatic |
| Email when a check finds something new | – | – | – | ✓ |
A few more answers
What's the difference between the Full Scan and the Launch Audit?
The Full Scan ($19) checks your live site from the outside with 40+ deep checks and gives you a complete fix guide. The Launch Audit ($79) adds what an outside scan can't see: seven specialist AI reviews of the code in your GitHub repo (access control, logins, payments and webhooks, input handling, secrets, AI features and abuse), a check of every library against known vulnerabilities (OSV), a secret scan of your full git history and realistic attack paths. A second, independent AI double-checks every finding. You get a one-page launch verdict, a prioritised fix plan with prompts for Cursor, Claude Code, Lovable, Bolt, Replit and v0, a detailed PDF and a free re-test.
Can you promise my app is 100% safe?
No, and nobody honestly can. We check for the problems our checks know how to find, at the moment of the scan. New kinds of problems appear, and your app changes every time you edit it. Some hosting companies also block parts of a security check, which is outside our control. Your report then shows everything we could check and says which part we could not reach.
What does Greenlit mean?
A verified full scan found no critical, high or medium issues. You get an embeddable badge with the scan date. It is not a guarantee your app is secure.
Is there a human reviewing my report?
No. Greenlight Scan is fully automated. In the Launch Audit, seven specialist AI reviews go through your code and a second, independent AI double-checks every finding before it reaches your report — anything it rejects is dropped, and anything it can't decide is listed separately as an unverified lead.
Latest guides
New on the blog
- Supabase publishable key vs secret key, explainedSupabase publishable vs secret key (and anon vs service_role): which one is safe in the browser, how to spot a leaked secret key, and how to rotate it.
- Stripe in an AI-built app: 7 checks before real moneyStripe security checklist for Lovable, Bolt, Cursor and v0 apps: server-side prices, webhook signatures, duplicate events and plan columns users can edit.
- Is my v0 app secure? Check before you share the linkIs my v0 app secure? Check NEXT_PUBLIC_ variables, Vercel environment variables, deployment protection, API routes and your database before you share your v0 link.
- Is my Replit app secure? Check before you share the linkIs my Replit app secure? Check Secrets, public repls, exposed files, open databases and API routes before you share your .replit.app link.
Secure your app, tool by tool
- How to secure your Lovable app before launchFour things to check in a Lovable app (Supabase tables, keys in the browser, edge functions and storage buckets) and the prompts to fix each one.
- How to secure your Bolt app before you share the linkBolt apps run in the browser and often call paid APIs directly. Learn how to move keys server-side, lock your database and publish safely.
- Cursor security: secrets, hooks and the app you shipKeep keys out of Cursor's reach (.cursorignore, .env, hooks, MCP tokens) and check the app it built for leaked keys and open data. Free scan.
- How to secure a Replit app (Secrets, public repls and databases)Use Replit Secrets properly, avoid exposing files through your deployment, and keep your database and API routes closed to strangers.
Ready to check your app?
The free scan takes about 2 minutes. No card, no account.
Scan my app free