Core Concepts

Extralt organizes extraction around three entities. Understanding how they relate is key to using the platform effectively.

Robot

A robot is an AI-generated crawler compiled to Rust, built for a specific host and country combination. Robots understand a site's structure: how to navigate pages, find products, and extract structured product data.

Creating a robot starts with a robot build. You provide a URL and country, and Extralt's AI analyzes the site, generates extraction logic, and compiles it. When the build succeeds, a robot appears in your robot list.

• One robot per host + country combination
• Robots are reusable across multiple runs
• If a site changes significantly, you can rebuild the robot

See Robots for build lifecycle and management.

Run

A run is an extraction job. You select a robot, provide start URLs, and optionally set a budget (maximum URLs to extract). The robot crawls the site and produces captures.

Run lifecycle:

1. Pending -- queued for execution
2. Running -- actively crawling and extracting
3. Completed -- finished successfully
4. Failed -- encountered an unrecoverable error
5. Stopped -- manually stopped by user

Each URL extracted in a run costs 1 credit.

See Running Extractions for run creation and monitoring.

Capture

A capture is a single extracted data record from a run. Each capture contains structured ecommerce product data.

{
  "_id": "abc123",
  "url": "https://example-store.com/products/sneakers",
  "data": {
    "title": "Air Max 90",
    "brand": "Nike",
    "description": "Classic sneaker...",
    "images": ["https://..."],
    "identifiers": {
      "gtin": "1234567890123",
      "mpn": "DN3707-100"
    },
    "variants": [
      {
        "option_values": ["10", "White"],
        "sku": "AM90-WHT-10",
        "price": { "amount": 130, "currency": "USD" },
        "availability": "in_stock"
      }
    ]
  }
}

See Working with Captures for accessing and exporting data.

How they relate

What's next

• Quickstart -- try it hands-on
• Account Setup -- create your org and get an API key
• How Extralt Works -- deeper dive into the architecture