AgentQL locates elements and data on live web pages from natural-language queries. Rather than matching XPath or DOM selectors, it analyses the structure of the page to find what the query describes, which lets the same query run across sites with similar content and keeps working as a page's markup changes. Queries also define the shape of the result, so extraction returns structured output, and transforms can be applied inside the query itself.
It works on public and authenticated pages, including dynamically generated content, and reaches pages behind a login by saving and reloading an authenticated session. Around the query language sit Python and JavaScript SDKs that drive Playwright for automation and testing, a REST endpoint that retrieves data from a URL without a browser, a browser extension for refining queries against a live page, and a hosted playground that exports Python scripts. Integrations cover agent frameworks and automation tools such as LangChain and Zapier, along with an MCP server.
Worked examples in both languages cover closing cookie dialogs and pop-ups, infinite scroll, pagination, form submission, site login, headless and remote browsers, stealth and human-like behavior against anti-bot measures, price comparison, map scraping to CSV, and news aggregation.
Features
- Natural-language selectors: elements and data are found by describing them, without XPath or CSS selectors
- Structured output: the shape of the query defines the shape of the returned data, with transforms applied in the query
- Cross-site queries: one query works across different sites with similar content
- Resilience to UI changes: queries keep resolving as a page's structure changes over time
- Playwright integration: the Python and JavaScript SDKs drive Playwright for automation and testing
- REST API: retrieve public-facing data from any URL without running a browser
- Authenticated and dynamic pages: works on private pages and content generated at runtime, with saved and reloaded sessions
- Browser debugger: an extension for optimizing queries in real time on a live page
- Playground: an online sandbox for trying queries and exporting Python scripts
- Framework integrations: connects to agent frameworks and automation tools including LangChain and Zapier, plus an MCP server
