Finance MCP Server

A Model Context Protocol (MCP) server that provides financial data for LLMs. This server specifically fetches financial statements (Income Statement, Balance Sheet, Cash Flow) for public companies.

Features

• Free to use: No API keys required.
• Structured Data: Returns data in consistent Pydantic models (JSON schema).
• Easy Integration: Works with any MCP-compliant client (Claude Desktop, Cursor, etc.).

Data Source

This package retrieves financial data from Yahoo Finance using the yfinance library. Note: This tool is for educational and research purposes. Please respect Yahoo Finance's terms of service.

Installation & Usage

Option 1: Quick Use with uvx (Recommended)

You can run this server directly without standard installation using uv (or uvx).

Add this to your MCP settings configuration (e.g., ~/.config/Claude/claude_desktop_config.json or your IDE's MCP config):

{
  "mcpServers": {
    "finance": {
      "command": "uvx",
      "args": [
        "finance-mcp-free"
      ]
    }
  }
}

Option 2: Local Development (Git Clone)

If you want to modify the code or run it locally:

1. Clone the repository:

   git clone https://github.com/ryar001/finance_mcp.git
   cd finance_mcp
   

2. Install dependencies:

   # Using uv (recommended)
   uv venv
   source .venv/bin/activate
   uv sync
   
   # Or using pip
   pip install -e .
   

3. Run the server:

   # Run directly
   finance-mcp
   
   # Or via fastmcp dev for auto-reload
   fastmcp dev finance_mcp/main_mcp.py:mcp_server
   

Data Models

The server returns data structured according to strict Pydantic models to ensure reliability for your LLM.

Example Return Format

All financial statements return a dictionary where keys are dates and values are the statement object.

Income Statement Example:

{
  "30/09/2023": {
    "ticker": "AAPL",
    "period": "30/09/2023",
    "total_revenue": 89498000000,
    "net_income": 22956000000,
    "cost_of_revenue": 49071000000,
    "gross_profit": 40427000000,
    "operating_expenses": 13458000000,
    "operating_income": 26969000000,
    ...
  }
}

The data is strictly typed to ensure your LLM can reliably query fields like total_revenue, net_income, etc.