MCP - Stock Analyzer

Model Context Protocol (MCP) server system for real-time stock market data scraping, storage and retrieval, analysis, and automated social media posting. Three specialized MCP servers that enable AI agents to collect, store and extract stock data, analyze market performance from yfinance, and post impressive market updates to X (Twitter)

Disclaimer

This project is for educational and research purposes only. Not financial advice. Use at your own risk when making investment decisions.😅

The project:

A demonstration of building MCP servers that integrate with Agents like Claude Desktop and VS Code (MCP Client). System's major 3 components:

1. Stock Scraping server: Real time stock data fetched from Yahoo Finance
2. Database server: Fetched data stored, retrieved, and analyzed using SQLite3.
3. X(Twitter) server: The analyzed data is curated into an impressive post and automatically posted as market updates on X (Twitter) using X APIs

So how does this work ? 🤔

The MCP allows the MCP Clients to use external tools. MCP servers described above shall expose the required tools that the AI/Client can call when in need.

1. Ask LLMs : "Get current price of Apple stock, and tweet about it" - very simple prompt.
2. Now, Claude shall call :
   • stock scraper tool to get Apple's current price
   • database tool to store the data, and retrieve
   • twitter tool to post a tweet with the information

The servers integrated into the AI now, the dropdown reveals the tools associated with them

The twitter post (https://x.com/schweigenSMB/status/1929122184720421040):

Setup

Prerequisites

• Python 3.10 or higher
• Twitter Developer Account with API access (Basic tier required for posting) Consumer API key and secret, and Access token and secret
• Claude Desktop or VS Code with MCP support (I have used these, there other clients out there, room for exploration)

Installation

• Clone and setup

  git clone https://github.com/bmsvinci1729/mcp-stock-analyzer.git
  cd mcp-stock-analyzer
  
  # Create virtual environment
  python3 -m venv stock-mcp
  source stock-mcp/bin/activate  # On Windows: stock-mcp\Scripts\activate
  
  # Install dependencies
  pip install mcp yfinance tweepy python-dotenv asyncio
  Well its age of "uv" too!!
  
  

• Configure your Environment variables Create a .env file in the project root:

  # Twitter API Credentials
  TWITTER_CONSUMER_KEY=your_consumer_key_here
  TWITTER_CONSUMER_SECRET=your_consumer_secret_here
  TWITTER_ACCESS_TOKEN=your_access_token_here
  TWITTER_ACCESS_TOKEN_SECRET=your_access_token_secret_here
  TWITTER_BEARER_TOKEN=your_bearer_token_here
  
  # Database Configuration
  DATABASE_PATH=/path/to/your/project/database/stocks.db
  
  

  Caution: Don't commit this file to your github, if you are experimenting or modifying further

• Now a configuration for our Agents/Clients, here is the claude_desktop_config.json: Claude Desktop -> File -> Settings -> Developer -> Edit config

  {
    "mcpServers": {
      "stock-scraping": {
        "command": "/path/to/your/project/stock-mcp/bin/python",
        "args": ["/path/to/your/project/stock_scraping_mcp_server.py"]
      },
      "stock-database": {
        "command": "/path/to/your/project/stock-mcp/bin/python",
        "args": ["/path/to/your/project/database_server.py"]
      },
      "stock-twitter": {
        "command": "/path/to/your/project/stock-mcp/bin/python",
        "args": ["/path/to/your/project/twitter_mcp_server.py"]
      }
    }
  }
  
  

• Begin your experiments (interacting):

`    "Get prices for Apple, Google, Tesla and Microsoft"

  "Store the current NVIDIA stock price in the database"
  "Show me Apple stock data from the last hour"
  "Get the top 5 performing stocks from today"
  "Post a tweet about the current market performance"

  "Analyze the top 10 tech stocks, identify the biggest movers, and post a professional market update tweet"

Learn more and dive deeper (resources)

• MCP
• MCP lecture by Anthropic
• X API
• Claude Desktop for Linux - Calmly do follow the instructions
• I did get several issues related to X API's 403 Forbidden error, do ensure token_ids and secrets are correct, sometimes a rate-limit issue.
• Maintain tweet size under 280 characters (Can be configured)
• Does this end here? No, the MCP lecture above discusses next steps and future of MCP which we must look forward towards
• My code could have been more modularized, and defined each try catch blocks in separate files, more modifications and other features like search tweets, adding some more analytic tools, scraping more stocks can be few improvements