Let your coding agent check your live app for leaks
LeakPatrol has an MCP server, so Claude Code, Cursor and other AI coding agents can run a security scan on your live site themselves, for example right after a deploy. The agent gets the grade, every problem and the launch checklist back, and can start fixing straight away. It's free, read-only and needs no API key.
Add it in one line
Claude Code
claude mcp add --transport http leakpatrol https://www.getleakpatrol.com/mcp
Cursor
Add this to ~/.cursor/mcp.json (or .cursor/mcp.json in your project):
{
"mcpServers": {
"leakpatrol": { "url": "https://www.getleakpatrol.com/mcp" }
}
}
VS Code, Windsurf and other MCP clients
Add a remote (Streamable HTTP) server with the URL https://www.getleakpatrol.com/mcp. There's no login or key.
Then just ask
- "Scan myapp.lovable.app with LeakPatrol and fix anything critical."
- "I just deployed. Check the live site for leaked keys before I share it."
- "Run the LeakPatrol launch checklist on my site and fix the link preview."
What the agent gets back
One tool, scan_website, which runs the same read-only checks as the scan on our home page:
- secret API keys in your JavaScript (OpenAI, Anthropic, Stripe, AWS, Supabase service keys and more), always masked
- public
.envand.gitfiles and source maps - paid features switched on from the browser, and AI providers called straight from the browser
- missing security headers and HTTPS problems
- the launch checklist: page title, link preview, sitemap, analytics, your own domain, DMARC
It returns an A–F grade, every issue, one worked fix and the launch checklist, as text and as structured data. The full report with a fix prompt for every issue is linked at the end.
Limits and safety
- Read-only: it fetches only what a browser would, plus a short list of files that should never be public. It never logs in, submits forms or changes anything.
- 5 scans per hour per connection; a site's result is reused for 10 minutes.
- Only scan sites you own or have permission to test.
Check your app in a minute
Free, read-only, no sign-up. Every problem comes with a plain-English fix and a prompt for your AI tool.
Get the 10-point checklist
Every check we run on AI-built apps, with how to do each one yourself. Straight to your inbox.
Plus an email when automatic re-checks launch, and the odd tip. Unsubscribe any time.
Questions
Does LeakPatrol have an MCP server?
Yes. Add https://www.getleakpatrol.com/mcp as a remote (Streamable HTTP) MCP server in Claude Code, Cursor, VS Code or Windsurf. It offers one read-only tool, scan_website, and needs no API key.
How do I add LeakPatrol to Claude Code?
Run: claude mcp add --transport http leakpatrol https://www.getleakpatrol.com/mcp. Then ask Claude Code to scan your site with LeakPatrol.
Is it safe to let an agent run security scans?
The scan is read-only. It only downloads what any visitor's browser would, plus a short list of files that should never be public, and any key it finds is masked.
More guides
- Supabase RLS policy generator (free)
- Is this API key safe in my frontend? Free key checker
- Supabase's October 30 change: what it fixes for your app's security, and what it doesn't
- Lovable security check: is your Lovable app safe?
- Bolt.new security check: is your Bolt app safe?
- Launch checklist for apps built with Lovable, Bolt and Cursor
- Supabase security check: is your anon key safe? (RLS checker)
- How to check if your API key is exposed on your website
- Are VITE_ env variables secret? How to hide API keys
- Stop users giving themselves your paid plan (Supabase)