MCP Agents

A powerful AI agents platform that enables automated interactions with various services and APIs, including web browsing, stock market data retrieval, and more.

Architecture

MCP Server Components

• Core Server:

  • Handles agent coordination and task management
  • Implements async communication patterns
  • Manages service integrations

• Service Integrations:

  1. Browser MCP:
     • Playwright-based web automation
     • Multi-tab browser management
     • Screenshot and PDF capabilities
     • Form interaction handling
  2. Finance MCP:
     • Stock market data integration (yfinance)
     • Historical and real-time data processing
  3. Airbnb MCP:
     • Property search and filtering
     • Listing details retrieval
     • Booking information handling
  4. Search MCP:
     • DuckDuckGo integration
     • Web search capabilities

Agent System

• MCPAgent Class:

  • Manages conversation state
  • Handles tool selection and execution
  • Implements memory management
  • Coordinates with LLM for decision making

• MCPClient Class:

  • Manages service connections
  • Handles configuration loading
  • Implements session management
  • Provides unified API access

Features

• Web Browsing Automation: Automated browser control using Playwright

  • Navigate websites
  • Handle forms and interactions
  • Take screenshots
  • Manage multiple tabs
  • Handle cookies and dialogs

• Financial Data Integration:

  • Stock market data retrieval using yfinance
  • Historical price data
  • Real-time market information

• Airbnb Integration:

  • Search for listings with customizable filters
  • Get detailed property information
  • Support for:
    • Check-in/check-out dates
    • Number of guests (adults, children, infants)
    • Price ranges
    • Location-based search
    • Pet-friendly options

• Search Capabilities:

  • Web search functionality
  • DuckDuckGo integration for privacy-focused searches

Prerequisites

• Python 3.x
• Virtual Environment (recommended)
• Node.js and NPX (for Playwright and browser automation)
• Groq API key (for LLM integration)

Setup

1. Clone the repository:

git clone <repository-url>
cd mcp_agents

2. Create and activate a virtual environment:

python -m venv .venv
source .venv/bin/activate  # On Unix/macOS
# or
.venv\Scripts\activate  # On Windows

3. Install dependencies:

pip install -r requirements.txt

4. Install Playwright using NPX:

# Install Playwright browsers using NPX
npx playwright install --with-deps

# For specific browsers only
npx playwright install chromium
npx playwright install firefox
npx playwright install webkit

5. Configure environment:

# Create .env file
touch .env

# Add required API keys
echo "GROQ_API_KEY=your_groq_api_key" >> .env

Development Tools

NPX Commands

The project uses NPX for various development and testing tasks:

# Run Playwright tests
npx playwright test

# Run Playwright in debug mode
npx playwright test --debug

# Show Playwright test report
npx playwright show-report

# Start Playwright Inspector
npx playwright codegen

Usage

Initialize MCP Client

from mcp_use import MCPClient, MCPAgent
from langchain_groq import ChatGroq

# Initialize client with configuration
client = MCPClient.from_config_file("browser_mcp.json")

# Initialize LLM
llm = ChatGroq(model="qwen-qwq-32b")

# Create agent
agent = MCPAgent(
    llm=llm,
    client=client,
    max_steps=15,
    memory_enabled=True
)

Web Automation

# Example of browser automation
await page.goto('https://www.example.com')
await page.click('button#submit')
await page.type('input#search', 'query')

Stock Market Data

# Example of retrieving stock data
stock_history = get_stock_history(
    symbol="AAPL",
    period="1mo",
    interval="1d"
)

Airbnb Search

# Example of searching Airbnb listings
airbnb_results = airbnb_search(
    location="Miami, FL",
    checkin="2024-06-01",
    checkout="2024-06-07",
    adults=2,
    children=0,
    max_price=300
)

# Get detailed information about a specific listing
listing_details = airbnb_listing_details(
    id="listing_id",
    checkin="2024-06-01",
    checkout="2024-06-07",
    adults=2
)

Web Search

# Example of performing a web search
search_results = duckduckgo_search(
    query="your search query",
    safe_search="moderate"
)

Project Structure

mcp_agents/
├── .venv/                 # Virtual environment
├── src/                   # Source code
│   ├── browser/          # Browser automation modules
│   ├── finance/          # Financial data modules
│   ├── airbnb/           # Airbnb integration modules
│   └── search/           # Search functionality modules
├── tests/                # Test files
│   └── playwright/       # Playwright test scripts
├── playwright.config.js  # Playwright configuration
├── browser_mcp.json      # Browser MCP configuration
├── package.json          # Node.js dependencies
├── requirements.txt      # Python dependencies
└── README.md            # This file

Configuration

browser_mcp.json

{
  "browser": {
    "headless": true,
    "viewport": {
      "width": 1280,
      "height": 720
    }
  },
  "services": {
    "airbnb": {
      "enabled": true
    },
    "finance": {
      "enabled": true
    },
    "search": {
      "enabled": true
    }
  }
}

Contributing

1. Fork the repository
2. Create a feature branch
3. Commit your changes
4. Push to the branch
5. Create a Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Playwright for web automation
• yfinance for financial data
• DuckDuckGo for search functionality
• Groq for LLM capabilities