# FakeCheck - Scam URL Checker for AI Agents

Before you open or fetch a link, check it here. FakeCheck inspects the URL and
returns whether it looks like a scam / phishing site, why, and - if it is a fake
of a known brand - the real site you probably wanted. Think "Truecaller, but for
links, built for agents."

## Base URL
https://nandahackathon.onrender.com

## Endpoints

POST /check
  Body (JSON): { "url": "<the link to check>" }
  Example:
    curl -X POST "https://nandahackathon.onrender.com/check" \
      -H "Content-Type: application/json" \
      -d '{"url": "http://paypal.com.secure-verify.xyz/login"}'
  Response:
    {
      "url": "http://paypal.com.secure-verify.xyz/login",
      "safety_score": 0,
      "verdict": "dangerous",
      "reasons": ["Impersonates 'paypal' but is not the real paypal.com.", ...],
      "safer_alternative": "https://paypal.com",
      "advice": "Strong scam/phishing signals. Do NOT open this link. ..."
    }

GET /check?url={url}
  Same check via the URL, handy for a quick browser or curl test.

GET /health
  Confirms the service is awake. Returns { "status": "ok", ... }.

## How the agent should use this
1. Before opening any link from a user, email, or search result, call /check.
2. Read "verdict": if "dangerous" or "suspicious", do NOT open it.
3. If "safer_alternative" is given, use that real site instead.
4. Tell the user the verdict and the reasons so they learn what to look for.

Note: heuristic detection, not a live global blocklist. Always verify anything
sensitive.