Your agent sends a URL and either an AgentQL query or natural language prompt to AgentQL's REST API. AgentQL's AI analyzes the page structure, identifies the requested elements semantically, and returns structured JSON data. The agent then formats this data conversationally for the customer.

Yes. AgentQL can create remote browser sessions that fully render JavaScript-heavy pages before extraction. For pages with dynamic content, infinite scroll, or single-page applications, the agent requests a browser session, waits for rendering, then queries the fully loaded page.

AgentQL's queries are self-healing. Instead of relying on brittle CSS selectors or XPath, AgentQL uses semantic understanding to find elements by meaning. If a website redesigns, queries like 'get product prices' continue working because AgentQL identifies prices by context, not DOM position.

Extracted data is used only for the current conversation and not permanently stored by Tars. When AgentQL returns structured data, your agent presents it to the customer and the response is logged as part of the conversation transcript, but raw extraction data is not separately cached or retained.

You can use AgentQL's query language with structured syntax like '{ products[] { name price } }' or natural language prompts like 'List all product names and their prices.' Natural language is easier for dynamic queries, while AgentQL syntax offers precise control over output structure.

AgentQL provides a free tier with initial API calls, then charges per query after. Your agent can check usage via the Get Usage endpoint to monitor consumption. Configure your agent with query limits or approval workflows for high-volume extraction scenarios.

The Tars integration supports public webpage extraction. For authenticated content, you would need to provide raw HTML directly to the query endpoint rather than a URL. AgentQL can parse HTML strings, enabling extraction from authenticated pages if your system captures the rendered HTML.

Custom scrapers require CSS or XPath selectors that break when sites change. AgentQL uses AI to understand page semantics, making queries resilient across site redesigns. You describe what data you want in natural language, not where it lives in the DOM. No scraper maintenance, no selector updates.