📊 📈 Stock & Crypto MCP Server with Local AI Agent

A modular, agent-based system for real-time stock prices, cryptocurrency news, and financial analysis using MCP (Multi-Component Protocol) servers, LangGraph agents, and LLMs (Ollama + Qwen3).

📌 Features

• 🔍 Stock Price Retrieval — Query real-time stock data (e.g., AAPL, TSLA) using a custom Yahoo Finance MCP server.
• 📰 Crypto News Aggregation — Stream cryptocurrency news headlines using a separate microservice.
• 🤖 LangGraph Agent Execution — A ReAct-style agent powered by LangGraph makes decisions about which tools to call based on context.
• 🧩 Tool Integration via MCP — Add more tools easily with the MCP protocol. Tools are separate microservices that expose LangChain-compatible interfaces.
• 🧠 LLM Backend with Ollama — Uses Qwen 1.7B running locally through Ollama for fast, local reasoning and tool selection.

🏗️ Architecture

• Agent Framework: LangGraph
• LLM Provider: Ollama (qwen3:1.7b)
• Tool Communication Layer: LangChain MCP Adapter
• Tool Servers: Custom Python servers (e.g., crypto_news.py, stock_news.py)

🚀 Getting Started

1. Clone the repository

git clone https://github.com/ChabchoubMedAziz/mcp-crypto-stock-news.git
cd mcp-crypto-stock-news

2. Install dependencies

pip install -r requirements.txt

“2”. Get Your API Key and Set Environment

• Create a .env file with your credentials:

AUTH_KEY=your_auth_key

Sign up on https://cryptopanic.com/developers/api/ and get your own Auth keys.

4. Start the MCP tool servers manually

python client.py

The system will:

• Spin up MCP clients

• Load tool definitions

• Use the agent to respond to a prompt like: “Give me AAPL stock price”

🛠️ Tools

The follwing tool is exposed to MCP clients:

income_statement()

-> This tool returns the quarterly income statement for a given stock ticker.

stock_info

->This tool returns information about a given stock given it’s ticker.

stock_price

->This tool returns the last known price for a given stock ticker.

get_crypto_news()

->Fetch the latest cryptocurrency news from CryptoPanic API

5. Extending the System

Adding a new tool is easy:

Write a microservice tool using LangChain’s MCP decorators.

Add the new server config in main.py.

Restart the system.

Example client.py config:

server_configs = {
  "weather_tool": {
    "command": "uv",
    "args": ["run", "weather.py"],
    "transport": "stdio"
  }
}

📝 License

• This project is licensed under the MIT License. See the LICENSE file for details.
• Built with Model Context Protocol