🦜🔗 MCPFinanceStream

MCPFinanceStream is a Streamlit web application powered by LangChain and FastMCP. It combines GPT-4o with a finance-oriented toolset that lets you analyze stock data interactively using MCP agents via SSE (Server-Sent Events).

📦 Features

• 💬 Natural language queries using Azure OpenAI (GPT-4o)
• 🔌 MCP integration using fastmcp and langchain-mcp-adapters
• 📈 Real-time stock data with yFinance
• 📊 Automatic data analysis using quarterly and annual financials
• 🚀 SSE-based architecture (no CLI or subprocess launch needed)

🏗️ Requirements

Make sure the following are installed (you can also use the requirements.txt):

streamlit
openai
langchain
langchain-openai
yfinance
pandas
langchain-mcp-adapters
langgraph
fastmcp

Python 3.9+ is recommended.

🚀 How to Run

Step 1: Start the MCP Server

python mcp_server.py

This starts your SSE-enabled server at http://localhost:8002/sse.

Step 2: Run the Streamlit App

streamlit run MCPFinanceStreamlit_app.py

Then open http://localhost:8501 in your browser.

📘 Example Query

Analyze how revenue of MSFT is changing over time.

The agent will call tools like fetch_quarterly_financials and fetch_annual_financials and generate a summary based on Microsoft's historical revenue.

⚙️ MCP Tools

• fetch_stock_info(symbol: str)
• fetch_quarterly_financials(symbol: str)
• fetch_annual_financials(symbol: str)

You can expand the toolset by modifying mcp_server.py.

🤖 Architecture

[Streamlit UI]
    ↓
[LangChain Agent (GPT-4o)]
    ↓
[MultiServerMCPClient → SSE]
    ↓
[FastMCP Server → yFinance]

🙋 FAQ

Q: What if I see raw ToolMessage or debug info? A: You’re seeing the full agent trace. We now extract only the final assistant message for a clean output.

Q: Does this use subprocesses? A: No! It's powered by SSE (asyncio) for smoother integration.