Page Agent

Page Agent (PageAgent) is an open-source (MIT) framework from Alibaba that puts a GUI agent directly inside any web page. With a single in-page JavaScript include, or an npm install, any web application gets its own AI agent that users can control with natural-language commands such as 'Click the login button.' Unlike most web-automation approaches, Page Agent needs no browser extension, no Python, and no headless browser, and it does not rely on screenshots or multi-modal LLMs. Instead it uses text-based DOM manipulation, so it works with most mainstream text models, including locally deployed ones, and requires no special permissions. This makes it lightweight and easy to embed. For builders and product teams, the primary use cases are shipping a SaaS AI copilot inside your product in just a few lines of code without a backend rewrite, smart form filling that turns 20-click workflows in ERP, CRM, and admin systems into a single sentence, and accessibility, letting users operate any web app through natural language, voice commands, or screen readers. Developers bring their own LLM by passing a model name, base URL, and API key (the docs show examples with Alibaba's Qwen via the DashScope compatible endpoint), and there is a free demo LLM for technical evaluation. An optional Chrome extension extends an agent's reach across multiple browser tabs for multi-page tasks. Page Agent also ships an MCP Server (Beta) that lets external agent clients control the browser via the Model Context Protocol, so your own agents can drive the page from outside. The project builds on the open-source browser-use project and is designed for client-side web enhancement rather than server-side automation. It is distributed on npm with TypeScript support and a Chrome Web Store extension.