Verified project record

alibaba/page-agent

An embeddable, in-page JavaScript GUI agent that enables natural language control of web interfaces without browser extensions, backend rewrites, or headless browsers. It processes the DOM as text, avoiding the need for screenshots or multi-modal models, and is designed for client-side web enhancement.

Project overview

Operates directly inside a target web page using a single script tag, offering a lower-overhead approach to web interaction compared to headless browsers or extensions.

Project type: MCP · AI Agent
Use cases: Automation
Deployment: Refer to project documentation
License: MIT

Best for

Developers needing to add natural language interface control into specific web pages without modifying the backend or requiring headless browser infrastructure.
Projects where an external AI client needs to control an embedded web agent via Model Context Protocol.

Key capabilities

Embeds a GUI agent directly into a web page using JavaScript to control web interfaces with natural language.
Interacts with the webpage through text-based DOM processing without relying on screenshots or multi-modal models.
Provides an optional Model Context Protocol (MCP) Server to control the agent from external clients.
Offers an optional Chrome extension to enable multi-page task automation across browser tabs.

Limitations and risks

Without the optional Chrome extension, the agent's scope is limited to a single page.
The free Demo LLM CDN is provided for technical evaluation only and may have specific usage terms; production use requires bringing your own LLM credentials.

Getting started

Setup is rated as easy. Developers can install via NPM or use a CDN script tag. The first success path involves installing the package, importing it, and instantiating PageAgent while providing credentials for a required external or paid LLM service.

Alternatives and comparisons

Controls browsers using a GUI agent, DOM, or a hybrid strategy using Vision-Language Models to recognize screen content.
An open-source Chromium fork that runs locally, preserving user logins and privacy while allowing AI tools to execute real web tasks.
Exposes Chrome browser functionality directly to AI assistants, enabling complex browser automation using the user's daily browser environment.

Project comparisons

Evidence and sources

GitHub project description: JavaScript in-page GUI agent. Control web interfaces with natural language.
README: - **🎯 Easy integration** - No need for `browser extension` / `python` / `headless browser`. - Just in-page javascript. Everything happens in your web page. - **📖 Text-based DOM ma…
README: SaaS AI Copilot — Ship an AI copilot in your product in lines of code. No backend rewrite.
README: import { PageAgent } from 'page-agent' const agent = new PageAgent({ model: 'qwen3.5-plus',
README: Fastest way to try PageAgent with our free Demo LLM:

Reviewed at: 2026-07-17T10:25:46Z

Open original GitHub project