Preventionlabs

ee what's wrong with a live website from the outside, with proof: uptime, HTTPS, DNS, broken pages, browser errors, leaked files, security headers, search visibility, speed and email setup. Read-only, and only on websites you've proven are yours. First 2 reports per site free, then A$2 a report.

Hosted MCP Server

npx add-mcp 'https://app.preventionlabs.com.au/mcp'

Installs into Claude Code, Codex, Cursor and more

Documentation

AI agents work from code. They can't see what visitors see: the broken page, the failed redirect, the certificate warning. That gap is invisible until a customer finds it.

Ask your agent to call Preventionlabs after it deploys. The report lands in your session: what's wrong, where, and what to do. See everything we check at /checks.

Connect it to your AI in a minute

No key, nothing to install. Add Preventionlabs as a custom connector with this address, sign in, and press Allow:

https://app.preventionlabs.com.au/mcp
  • claude.ai (web, desktop and phone): Settings, Connectors, Add custom connector. Paste the address and connect.
  • ChatGPT (developer mode): Settings, Apps and connectors. Turn on developer mode, create a connector with the address, and choose OAuth.
  • Gemini (gemini.google.com, personal accounts): Settings, Connected Apps, Add a custom app. Paste the address. Gemini Apps activity must be on.

Then ask it, by name the first time:

Use Preventionlabs to check my website.

Naming Preventionlabs tells your AI to use it rather than search the web. It runs a report on your live site and answers from the proof. You need an account with your site proven, and you can disconnect the app any time under Integrations.

How it works

  1. **You connect it once.**Create an API key in the app and paste one line into your agent. That's the whole setup.
  2. **Your agent ships a change.**Then it runs a Preventionlabs report on your live site. It takes seconds.
  3. **You see what the world sees.**The report lands in your session: what's wrong, where, the evidence and what to do. If it's clean, the problem isn't on the outside.
  4. **A second report proves the fix.**Run the report again after a fix. It shows whether the problem is gone.

Each website's first 2 reports are free, then A$2 a report from your balance. Your agent can reread the last report any time without using one.

Claude Code, Cursor and other AI agents

For agents that run on your computer, use an API key instead.

  1. Create your account, add your site and prove it's yours. (Claude Code, Cursor, Claude Desktop and other apps on your computer.)
  2. Open Integrations → AI agents and create your API key.
  3. Add Preventionlabs to your agent. Claude Code:
    claude mcp add -s user preventionlabs -e PREVENTIONLABS_API_KEY=plk_your_key -- npx -y preventionlabs-mcp
    
    Cursor, Claude Desktop, Windsurf and other MCP apps:
    {
      "mcpServers": {
        "preventionlabs": {
          "command": "npx",
          "args": ["-y", "preventionlabs-mcp"],
          "env": { "PREVENTIONLABS_API_KEY": "plk_your_key" }
        }
      }
    }
    

Safe by design

  • Read-only. Your agent can check your site. It can't change anything through Preventionlabs.
  • Only your sites. Only sites you've proven are yours can be checked.
  • You control the key. Delete it any time. You're emailed whenever a new one is created.

Price

A$2 a report. No subscription. Your first 2 reports on each domain are free.

Top up from A$20 (10 reports). What's left after a report stays on your balance for the next one.