About this project
Page Agent is an open-source, MIT-licensed JavaScript library from Alibaba that embeds a GUI agent directly inside a web page. With a single script tag or npm install, developers can give any site an AI agent that operates via text-based DOM manipulation—no screenshots, multi-modal LLMs, headless browser, Python, or browser extension required.
Key features include easy in-page integration, support for bringing your own mainstream or locally deployed LLMs, an optional Chrome extension for multi-page tasks across tabs, and a beta MCP Server that allows external agent clients to control the browser.
Documented use cases cover SaaS AI copilots that can be shipped without a backend rewrite, smart form filling to reduce multi-step workflows in ERP/CRM/admin systems to a single sentence, accessibility improvements via natural language or voice commands, and multi-page agent scenarios.
Quick start options include a one-line CDN demo script (for technical evaluation with a free testing LLM) and a standard npm package where developers configure their own model, baseURL, API key, and language, then call methods like agent.execute('Click the login button'). The project is built in TypeScript, acknowledges derivation of DOM processing components and prompts from the browser-use project, and welcomes community contributions (though purely bot/AI-generated contributions are not accepted).
Comments
0 Rating appears after 10 ratings
Sign in to join the discussion.