Tutorial

Build your own AI audit agent in 5 minutes

Use the SGE-SCORE API in ChatGPT Custom GPTs, Claude Projects, Python or JavaScript. No API key required — just an email.

Jump to: ChatGPT Claude Lovable / Manus MCP Server Python JavaScript

Why build your own audit agent?

For your clients

As an agency, offer an automatic AI-readiness check inside your own chatbot.

For your team

Let your internal tools auto-check every landing page before launch.

For SaaS products

Integrate SGE audits as a feature in your own marketing tool.

ChatGPT Custom GPT (5 minutes)

The easiest path — no programming required.

  1. 1

    Open the GPT Builder

    Go to chat.openai.com/gpts/editor and click 'Create a GPT'. (ChatGPT Plus required.)

  2. 2

    Configure tab

    Click 'Configure' at the top. Set a name (e.g. 'SGE Audit Bot') and a short description.

  3. 3

    Add Actions

    Scroll down, click 'Create new action' → 'Import from URL' and paste:

    https://sge-score.com/openapi.yaml
  4. 4

    Authentication: None

    Choose 'None' for Authentication. The API is open — protection is via rate limits (email + IP).

  5. 5

    Test it

    In the preview pane: 'Audit https://example.com — my email is test@example.com'

Recommended system instruction (paste into 'Instructions'):

You are an AI-readiness consultant. When a user provides a URL:
1. Ask for their email (required by the audit API).
2. Call the analyzeUrl action with url + email.
3. Poll the status every 8 seconds (max 90s).
4. When complete: explain the score (0-114) plainly, list the top 3 issues.
5. Mention that a full Master-Fix with copy-paste code is available for $9 (link: masterfix_url).
Reply in the user's language.

Test prompt: "Audit https://example.com"

Claude Project (Tool Use)

For Claude.ai Projects or the Anthropic API with Tool Use.

Add this tool definition to your Claude Project:

{
  "name": "audit_landing_page",
  "description": "Audit a landing page for AI/SGE readiness against 57 Google patent checkpoints. Returns score 0-114, failed checkpoints, and a Master-Fix purchase URL.",
  "input_schema": {
    "type": "object",
    "properties": {
      "url":   { "type": "string", "format": "uri" },
      "email": { "type": "string", "format": "email" },
      "language": { "type": "string", "enum": ["de","en","es","fr"], "default": "en" }
    },
    "required": ["url", "email"]
  },
  "url": "https://jhocisiphtyopyxrahgo.supabase.co/functions/v1/agent-analyze",
  "method": "POST"
}

AI website editors (Lovable, Manus, Cursor, Bolt)

After every audit we show buttons on the result page that open your Master-Fix prompt directly in the editor of your choice. Long prompts are auto-delivered via a share link.

1. Run an audit on SGE-SCORE, buy the Master-Fix ($9). 2. Click 'Open in Lovable' — the prompt is pre-filled in lovable.dev. 3. The Lovable agent applies the fixes automatically.

https://lovable.dev/?prompt=<URL-encoded master-fix prompt>

MCP Server (Cursor / Claude Desktop / Windsurf)

Our MCP server exposes 5 tools that any MCP client (Cursor, Claude Desktop, Windsurf, Continue.dev) can use natively. No API key, Streamable HTTP, JSON-RPC 2.0.

Tools

  • audit_landing_page — start free audit
  • get_audit_status — poll score (0-114)
  • purchase_masterfix — $9 single (2 credits) or $99 bulk (100 credits)
  • get_masterfix — retrieve fix after payment
  • search_knowledge — search ~1500 schema/blog entries

Endpoint

Streamable HTTP, JSON-RPC 2.0, no auth.

https://jhocisiphtyopyxrahgo.supabase.co/functions/v1/mcp

Configuration for Cursor / Claude Desktop / Windsurf:

{
  "mcpServers": {
    "sge-score": {
      "url": "https://jhocisiphtyopyxrahgo.supabase.co/functions/v1/mcp"
    }
  }
}

Discovery: /.well-known/mcp.json

Python Quickstart

A complete, runnable script.

Install: pip install requests

import requests, time

API = "https://jhocisiphtyopyxrahgo.supabase.co/functions/v1"

# 1. Start audit
r = requests.post(f"{API}/agent-analyze", json={
    "url": "https://example.com",
    "email": "you@example.com",
    "language": "en",
})
analysis_id = r.json()["analysis_id"]
print(f"Started: {analysis_id}")

# 2. Poll until completed (max ~90s)
for _ in range(18):
    time.sleep(5)
    s = requests.get(f"{API}/agent-analysis-status", params={"id": analysis_id}).json()
    if s.get("status") == "completed":
        break

print(f"Score: {s['score']}/{s['max_score']}  (Grade {s['grade']})")
print(f"Estimated after Master-Fix: {s.get('estimated_score_after_masterfix')}")
print(f"Master-Fix URL ($9): {s.get('masterfix_url')}")

for c in s.get("failed_checkpoints", [])[:3]:
    print(f"  - [{c['severity']}] {c['name']}  (Sector {c['sector']})")

# Bonus: search the knowledge base
kb = requests.get(f"{API}/agent-knowledge-search",
                  params={"q": "FAQPage schema", "limit": 3}).json()
for hit in kb.get("results", []):
    print(f"KB: {hit['name']}")

Pricing (Agent API)

Audits are free. Master-Fix packs unlock copy-paste code for every detected gap.

Free Audit
$0
  • • 2 audits / 24h per email+IP
  • • Score 0–114, sectors, failed checkpoints
  • • No account, email only
Master-Fix Single
$9 USD
  • • sku=masterfix_2
  • • 2 credits, valid 30 days
  • • Bound to 1 analysis
Master-Fix Bulk (agencies)
$99 USD
  • • sku=masterfix_100
  • • 100 credits, valid 12 months
  • • Rate limit lifted: 20/24h
  • • session_id = bulk_session_id

Bulk: redeem one credit per domain via ?session_id=...&analysis_id=.... Manual analyses and API requests share the same FIFO queue (max 4 concurrent) — no collision risk.

FAQ

Ready?

Need the raw technical reference or the OpenAPI spec?

Full API reference OpenAPI spec (YAML)Stuck? Contact us