MCP Web Automation Agent

A powerful AI-driven web automation agent that lets users control browser actions using natural language commands. Built with Streamlit, OpenAI, and Puppeteer.

Features

• Parse and execute web automation instructions using LLMs
• Dynamic command interpretation (e.g., "Search for flights to New York")
• Interface built with Streamlit for easy interaction
• Headless browser automation powered by Puppeteer

Tech Stack

• Python 3.9+ – Application logic and LLM integration
• Streamlit – Frontend UI for user interaction
• OpenAI API – Natural language understanding and command generation
• Node.js + Puppeteer – Browser automation layer
• YAML – Configuration management
• Logging – Local logging for troubleshooting and audit

How It Works

1. User Input: A natural language instruction is provided through the Streamlit interface.
2. LLM Parsing: The input is processed using OpenAI's API to generate a structured response.
3. Browser Automation: The agent sends this structured command to a Puppeteer-powered backend to simulate browser activity.

Project Structure

MCP--WebAutomation-Agent/
│
├── main.py                  # Core logic and Streamlit interface
├── requirements.txt         # Required Python packages
├── mcp_agent.config.yaml    # Configurations for the agent
└── logs/                    # Runtime logs

Getting Started

1. Clone the repo:

   git clone https://github.com/rijul21/MCP--WebAutomation-Agent.git
   cd MCP--WebAutomation-Agent

2. Install dependencies:

   pip install -r requirements.txt

3. Run the app:

   streamlit run main.py

Requirements

• Python 3.9+
• Node.js (for Puppeteer)
• OpenAI API Key

Notes

• Puppeteer should be configured separately to interact with the automation commands.
• All outputs and errors are logged under the logs/ directory.

License

This project is licensed under the MIT License.