</>MCP Agents Market
MCP Server

page-agent

by alibaba28.8kTypeScriptUpdated 2026-08-17

JavaScript in-page GUI agent. Control web interfaces with natural language.

Claude Desktop

Page Agent MCP server enables AI assistants to control web page interfaces through natural language commands using text-based DOM manipulation. Developed by Alibaba, this JavaScript-based solution runs directly in web pages without requiring browser extensions, Python, or headless browsers for basic operation. It supports most mainstream language models including locally deployed ones, and provides an optional MCP server interface for external agent clients to orchestrate browser automation tasks. The tool is particularly useful for building AI copilots into SaaS products and automating complex multi-step web workflows.

Key Features

In-page JavaScript execution without requiring browser extensions or headless browsers for core functionality
Text-based DOM manipulation that works without screenshots or multimodal LLM capabilities
Compatibility with mainstream language models including locally deployed options
MCP server interface (Beta) for external agent client control
Optional Chrome extension for multi-page automation tasks across browser tabs
Single-script integration for adding AI agent capabilities to any webpage
Support for natural language commands to execute complex web interactions
MIT licensed open-source project with TypeScript implementation

Use Cases

  • 01Building AI copilots into SaaS products without backend rewrites
  • 02Automating form-filling workflows in ERP, CRM, and administrative systems
  • 03Creating accessibility solutions with voice commands and natural language interfaces
  • 04Enabling multi-page browser automation through the Chrome extension
  • 05Allowing MCP-compatible agent clients to control browser sessions
  • 06Simplifying complex multi-click workflows into single natural language instructions

Related MCP Servers

View more

page-agent — FAQ

What is the Page Agent MCP server?+

Page Agent MCP server is a browser automation tool that lets AI assistants control web interfaces using natural language commands. It runs as in-page JavaScript and optionally provides an MCP server interface for external agent clients to orchestrate browser tasks.

How do I install Page Agent with my own LLM?+

Install via npm with 'npm install page-agent', then instantiate with your LLM credentials: new PageAgent({ model: 'your-model', baseURL: 'your-api-url', apiKey: 'YOUR_API_KEY' }). For quick testing, you can use the CDN script with the demo LLM included.

Which AI clients work with Page Agent MCP server?+

Page Agent's MCP server (Beta) is designed to work with MCP-compatible agent clients like Claude Desktop and other tools supporting the Model Context Protocol. The core library also integrates directly into web applications via JavaScript.

Do I need API keys to use Page Agent?+

For production use, you need API keys for your chosen LLM provider (such as Alibaba's DashScope for Qwen models, or other compatible providers). A free testing API is available via the demo CDN for technical evaluation only.

Is Page Agent free and open source?+

Yes, Page Agent is released under the MIT License and is completely open source. The core library is free to use, though you'll need your own LLM API credentials for production deployments.

What are the prerequisites for running Page Agent?+

You need a modern web browser and access to a compatible language model API (Qwen, OpenAI, Anthropic, or others listed in the documentation). No browser extensions, Python, or headless browsers are required for basic in-page functionality, though an optional Chrome extension is available for multi-page tasks.

How do I install page-agent?+

Open the source repository on GitHub and follow its README. page-agent is a mcp server — MCP Agents Market links you directly to the official repo.

Is page-agent free?+

page-agent is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.

Related searches