Your First Extraction

This guide walks you through a complete extraction from start to finish.

Prerequisites

• An Ev3ry account
• A website URL you want to extract from

1. Add a website

From the dashboard, click Add Website. Enter:

• URL — the page the agent should start from
• Name — a label for your own reference
• Description (optional) — describe what you want, e.g. "Extract upcoming sports events including name, date, venue, and ticket price"

The description guides the agent during exploration. The more specific, the better.

2. Create a schema

Go to your website's detail page and click New Schema (or open the global Schemas library to reuse an existing one).

Define the JSON Schema for the data you want. For a first extraction, a simple object schema works well:

{
  "type": "array",
  "items": {
    "type": "object",
    "properties": {
      "event_name": { "type": "string" },
      "date":       { "type": "string" },
      "venue":      { "type": "string" },
      "price":      { "type": "number" }
    }
  }
}

Tip: Start with 3–5 fields. You can always refine after seeing what the agent returns.

3. Start a run

Click Run. Choose your browser connection method:

• Built-in browser — runs locally on your machine, best for sensitive data or anti-bot protection
• Cloud browser — no setup needed if you have an API key configured in Settings

The agent opens a browser and begins working. The live view shows you the browser in real time — you can watch it navigate, click, and extract.

4. Review the results

When the run completes, the extracted data appears as a table. From here:

• Inspect rows for accuracy
• Download as JSON or CSV
• Save as a playbook for future reuse

Troubleshooting

Empty results — the page may load data dynamically. Add "scroll down to load all content" to the website description and re-run.

Wrong data — refine the schema field names and add a description to each property, e.g. "description": "the full event name including series title". The agent uses these hints.

Login wall — the agent landed on a login page. Add a saved login under the website's settings. See Authenticated Pages.

Slow run — first runs explore the page and write extraction scripts. Once you save a playbook, subsequent runs are significantly faster.