AgentQL

AgentQL, made by TinyFish, is a query language and parser that lets AI agents and LLM applications extract structured data from web pages and PDFs without writing brittle XPath or CSS selectors. Instead of targeting a specific element's exact position in the page structure, queries describe what to find in natural terms, and AgentQL's AI-powered matching stays robust even as a page's layout changes. Who it's for: developers building AI agents or automation that needs to reliably read data off real websites — a task that's traditionally fragile, since a normal site redesign can silently break a hand-written CSS selector. Strengths: ships with Python and JavaScript SDKs with Playwright integration, a browser-based debugger, a Chrome extension for interactive query-building, and a REST API for browserless extraction — a genuinely complete toolkit rather than just a parsing library. "Self-healing" queries that adapt to layout changes directly address the most common failure mode of traditional scraping. Weaknesses and limitations: like any tool that interacts with third-party websites, reliability still depends on what the target site allows (rate limits, anti-bot measures) — AgentQL reduces layout-change fragility specifically, not every failure mode of web automation. The free Starter tier is capped at 50 API calls a month, so any real production use needs a paid plan. Real-world scenario: an AI agent needs to pull current pricing or availability data from a set of retailer websites on a schedule, and the team doesn't want that automation to silently break every time one of those sites tweaks its page layout.