LangChain MCP Multi-Server Integration

A comprehensive Model Context Protocol (MCP) implementation that integrates multiple data sources and services with LangChain agents for intelligent data processing and storage.

🚀 Features

• Multi-Server Architecture: Weather, Stocks, MongoDB, and Elasticsearch servers
• AI Agent Integration: LangChain agent that intelligently uses all available tools
• Automatic Data Storage: Conversations and data automatically stored in Elasticsearch
• Real-time Data: Live weather alerts and stock portfolio data
• Secure Configuration: Credentials managed via .netrc files

📋 Prerequisites

• Python 3.8+
• Elasticsearch running on localhost:9200
• MongoDB running on localhost:27017 (optional)
• Kite Connect API account for stock data
• LLM API access (OpenAI-compatible endpoint)

🛠️ Installation

1. Clone the repository:

git clone https://github.com/vamsivenkat987/langchain_mcp_server.git
cd langchain_mcp_server/mcp

2. Install dependencies:

pip install langchain-mcp-adapters langchain-openai elasticsearch pymongo kiteconnect requests

3. Start required services:

# Start Elasticsearch
docker run -d -p 9200:9200 -e "discovery.type=single-node" elasticsearch:8.11.0

# Start MongoDB (optional)
docker run -d -p 27017:27017 mongo:latest

⚙️ Configuration

Create a .netrc file in the project directory:

machine llm.model
login your_api_key_here
password https://your-llm-endpoint-here

machine stocks.data
login your_kite_api_key
password your_kite_api_secret
account your_kite_access_token

🏃‍♂️ Usage

Quick Start

python3 mcp_client.py

Individual Server Testing

# Test weather server
python3 weather_server.py

# Test stock server
python3 stocks_server.py

# Test Elasticsearch server
python3 elastic_search_server.py

# Test MongoDB server
python3 mongo_server.py

🔧 Available Tools

Weather Server (weather_server.py)

• get_alerts(state) - Get active weather alerts for a US state
• Uses National Weather Service API

Stock Server (stocks_server.py)

• get_holdings() - Retrieve current stock portfolio from Kite Connect
• Requires valid Kite Connect session

Elasticsearch Server (elastic_search_server.py)

• add_document(document, index, doc_id) - Store documents in Elasticsearch
• Automatic timestamp and UUID generation
• JSON parsing for string inputs

MongoDB Server (mongo_server.py)

• add_conversation_data(user_content, ai_content) - Store conversations
• add_customer_data(name, age) - Store customer information

📁 Project Structure

mcp/
├── mcp_client.py              # Main LangChain agent client
├── weather_server.py          # Weather alerts MCP server
├── stocks_server.py           # Stock portfolio MCP server
├── elastic_search_server.py   # Elasticsearch MCP server
├── mongo_server.py            # MongoDB MCP server
├── test_elastic_search.py     # Elasticsearch testing utilities
├── test_kite.py              # Kite Connect authentication flow
├── test_mcp.py               # MCP server testing
├── .netrc                    # API credentials (create this)
├── oneEGaDCA.crt            # SSL certificate
└── README.md                 # This file

🤖 AI Agent Workflow

The LangChain agent follows this workflow:

1. Process User Query: Understand what the user is asking for
2. Call Appropriate Tools:
   • Weather queries → get_alerts(state)
   • Stock queries → get_holdings()
3. Provide Natural Response: Answer user's question in plain language
4. Store Data: Automatically save conversation and data to Elasticsearch

📊 Data Storage

Elasticsearch Indices

• conversations - User queries and AI responses
• weather - Weather alert data
• stocks - Stock portfolio information
• customer_data - Customer information

MongoDB Collections

• conversations - Chat history
• customers - Customer data

🔐 Security Features

• Credentials stored in .netrc files (not in code)
• SSL certificate support
• Environment variable configuration
• Secure API token handling

🔍 Example Queries

• "Get me the current weather alerts for New York"
• "What are my stock holdings?"
• "Is there any connection between my stock portfolio and weather in NY?"
• "Show me the weather in California and save the data"

🐛 Troubleshooting

Common Issues

1. Elasticsearch Connection Error

   • Ensure Elasticsearch is running on localhost:9200
   • Check if the service is accessible: curl http://localhost:9200

2. Kite Connect Authentication

   • Verify API credentials in .netrc
   • Check if access token is valid (tokens expire daily)

3. MCP Server Errors

   • Check server logs for detailed error messages
   • Ensure all dependencies are installed

4. LLM Timeout Errors

   • Increase timeout in mcp_client.py
   • Try using streaming instead of ainvoke

📝 API Documentation

Weather Server

• Endpoint: National Weather Service API
• Rate Limits: No authentication required
• Data Format: JSON weather alerts

Stock Server

• Endpoint: Kite Connect API
• Authentication: API key + secret + access token
• Rate Limits: As per Zerodha limits
• Data Format: Portfolio holdings with P&L

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

🆘 Support

For issues and questions:

• Create an issue on GitHub
• Check the troubleshooting section
• Review server logs for detailed error information

🔄 Version History

• v1.0.0 - Initial release with multi-server MCP integration
• v1.1.0 - Added automatic data storage and improved error handling
• v1.2.0 - Enhanced authentication and security features