Structured product data for AI agents. Product data where platform APIs don't reach.
Across 7 shopping verticals · Last updated: 2026-03-25
| Vertical | Pages | Success | Confidence |
|---|---|---|---|
| Fashion & Apparel | 35 | 91% | 0.85 |
| Health & Beauty | 22 | 86% | 0.84 |
| Electronics & Tech | 21 | 95% | 0.73 |
| Home & Furniture | 19 | 84% | 0.82 |
| Sports & Outdoors | 12 | 100% | 0.85 |
| Jewelry & Accessories | 4 | 100% | 0.95 |
| Food & Beverage | 3 | 100% | 0.86 |
Schema.org path: 0.95 confidence, near-instant · LLM fallback: 0.7+ confidence · Ground truth verified via Playwright
AI shopping agents need structured product data. ShopGraph extracts it from any page on the web.
<div class="product"> <span>Sony WH-1000XM5</span> <span class="price">$348</span> ...hundreds more lines... </div>
{
"name": "Sony WH-1000XM5",
"brand": "Sony",
"price": 348.00,
"currency": "USD",
"availability": "InStock",
"confidence": { "overall": 0.95 }
}
Two extraction modes, priced by complexity.
Full extraction with LLM fallback. Schema.org first, Gemini when markup is missing. All attributes, confidence scores, and images.
Schema.org and meta-tag extraction only. Fast, no LLM. Best for pages with good structured markup.
Cached results (within 24 hours) are free. Failed extractions are not charged.
Your agent sends any product URL to ShopGraph via MCP.
Schema.org parsing first. If markup is missing, Gemini LLM extracts from raw page content.
Clean data with confidence scores. Cached for 24 hours for free repeat access.
Every field includes a confidence score so your agent knows how much to trust it.
Any MCP-compatible client can connect to ShopGraph.
Connect any MCP-compatible client. Pay per call via Stripe Machine Payments Protocol.
Connect your MCP client:
{
"mcpServers": {
"shopgraph": {
"type": "url",
"url": "https://shopgraph.dev/mcp"
}
}
}
No API keys needed for discovery. Payment handled via Stripe MPP โ your agent pays per call using a Stripe-issued token. No subscriptions, no minimums.