AgentQL is an open-source web data extraction tool created by the American startup TinyFish. Instead of selectors tied to page structure such as CSS selectors or XPath, it uses natural-language queries describing the desired elements or data to locate them in HTML pages and return structured data. It was released in 2024.
Because a language model interprets the meaning of the page to find elements, queries keep working when layouts or CSS classes change. It offers Python and JavaScript SDKs, is used alongside browser automation tools such as Playwright, and is applied to web scraping, test automation, and building web agents.