Skip to main content
Glama
fritzprix

yfin-mcp

by fritzprix
README.md
# yfin-mcp - Yahoo Finance MCP Server

> **Enhanced fork** of [yahoo-finance-mcp](https://github.com/Alex2Yang97/yahoo-finance-mcp) by Alex2Yang97  
> With intelligent pagination, caching, and LLM-optimized responses

<div align="right">
  <a href="README.md">English</a> | <a href="README.zh.md">δΈ­ζ–‡</a>
</div>

A high-performance Model Context Protocol (MCP) server that provides comprehensive financial data from Yahoo Finance with **intelligent pagination**, **caching**, and **LLM-optimized responses**.

---

## πŸ™ Attribution & Motivation

### Original Work
This project is built upon the excellent foundation of [yahoo-finance-mcp](https://github.com/Alex2Yang97/yahoo-finance-mcp) created by **Alex2Yang97**. The original implementation provides a comprehensive set of tools for accessing Yahoo Finance data through the Model Context Protocol.

### Why These Enhancements?

While using the original implementation with LLM agents (Claude, ChatGPT, etc.), I encountered critical limitations:

**Problem 1: Context Window Overflow**
- Historical stock data with `period="max"` could return thousands of rows
- Option chains for popular stocks contained hundreds of contracts
- LLM context windows (typically 200K tokens) would overflow
- Responses would be truncated, losing critical data

**Problem 2: No Data Persistence**
- Large datasets couldn't be saved for offline analysis
- Repeated queries wasted API calls and time
- No way to export data for use in other tools

**Problem 3: Poor LLM Readability**
- JSON responses were hard for LLMs to parse when truncated
- No clear navigation guidance for paginated data
- Cache status was invisible to the LLM

### Solution: Pagination, Caching & Export

This fork adds three key enhancements:

1. **Token-Based Pagination** (6,000 token limit per page)
   - Prevents context window overflow
   - Clear navigation guidance for LLMs
   - Dynamic page sizing based on data complexity

2. **Intelligent Caching** (TTL: 5min-1hr)
   - Reduces redundant API calls by 50-80%
   - Sub-millisecond response times for cached data
   - Automatic cache invalidation based on data volatility

3. **JSON Export** (File Download)
   - Save full datasets for offline analysis
   - Export to Excel, databases, or other tools
   - Preserve complete data without pagination

### Credit Where Credit is Due

**Original Author**: Alex2Yang97 deserves full credit for:
- βœ… Complete Yahoo Finance API integration
- βœ… All 9 MCP tools implementation
- βœ… Robust error handling
- βœ… Comprehensive documentation

**This Fork Adds**: Pagination, caching, and export features to make the server production-ready for LLM agents handling large financial datasets.

## Demo

![MCP Demo](assets/demo.gif)

## MCP Tools

The server exposes the following tools through the Model Context Protocol:

### Stock Information

| Tool | Description |
|------|-------------|
| `get_historical_stock_prices` | Get historical OHLCV data for a stock with customizable period and interval |
| `get_stock_info` | Get comprehensive stock data including price, metrics, and company details |
| `get_yahoo_finance_news` | Get latest news articles for a stock |
| `get_stock_actions` | Get stock dividends and splits history |

### Financial Statements

| Tool | Description |
|------|-------------|
| `get_financial_statement` | Get income statement, balance sheet, or cash flow statement (annual/quarterly) |
| `get_holder_info` | Get major holders, institutional holders, mutual funds, or insider transactions |

### Options Data

| Tool | Description |
|------|-------------|
| `get_option_expiration_dates` | Get available options expiration dates |
| `get_option_chain` | Get options chain for a specific expiration date and type (calls/puts) |

### Analyst Information

| Tool | Description |
|------|-------------|
| `get_recommendations` | Get analyst recommendations or upgrades/downgrades history |

## Real-World Use Cases

With this MCP server, you can use Claude to:

### Stock Analysis

- **Price Analysis**: "Show me the historical stock prices for AAPL over the last 6 months with daily intervals."
- **Financial Health**: "Get the quarterly balance sheet for Microsoft."
- **Performance Metrics**: "What are the key financial metrics for Tesla from the stock info?"
- **Trend Analysis**: "Compare the quarterly income statements of Amazon and Google."
- **Cash Flow Analysis**: "Show me the annual cash flow statement for NVIDIA."

### Market Research

- **News Analysis**: "Get the latest news articles about Meta Platforms."
- **Institutional Activity**: "Show me the institutional holders of Apple stock."
- **Insider Trading**: "What are the recent insider transactions for Tesla?"
- **Options Analysis**: "Get the options chain for SPY with expiration date 2024-06-21 for calls."
- **Analyst Coverage**: "What are the analyst recommendations for Amazon over the last 3 months?"

### Investment Research

- "Create a comprehensive analysis of Microsoft's financial health using their latest quarterly financial statements."
- "Compare the dividend history and stock splits of Coca-Cola and PepsiCo."
- "Analyze the institutional ownership changes in Tesla over the past year."
- "Generate a report on the options market activity for Apple stock with expiration in 30 days."
- "Summarize the latest analyst upgrades and downgrades in the tech sector over the last 6 months."

## Requirements

- Python 3.11 or higher
- Dependencies as listed in `pyproject.toml`, including:
  - mcp
  - yfinance
  - pandas
  - pydantic
  - and other packages for data processing

## Installation

### From PyPI (Recommended)

Install the package directly from PyPI:

```bash
pip install yfin-mcp
```

### From Source

1. Clone this repository:
   ```bash
   git clone https://github.com/fritzprix/yahoo-finance-mcp.git
   cd yahoo-finance-mcp
   ```

2. Create and activate a virtual environment and install dependencies:
   ```bash
   uv venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate
   uv pip install -e .
   ```

## Usage

### Integration with Claude for Desktop

After installing the package, you can integrate it with Claude for Desktop:

1. **Install the package** (if not already installed):
   ```bash
   pip install yfin-mcp
   ```

2. **Configure Claude Desktop**:
   - MacOS: Edit `~/Library/Application Support/Claude/claude_desktop_config.json`
   - Windows: Edit `%APPDATA%\Claude\claude_desktop_config.json`

3. **Add the server configuration**:

   **Using uvx (Recommended - No installation needed)**:
   ```json
   {
     "mcpServers": {
       "yfinance": {
         "command": "uvx",
         "args": ["yfin-mcp"]
       }
     }
   }
   ```

   **Using Python directly (if installed via pip)**:
   ```json
   {
     "mcpServers": {
       "yfinance": {
         "command": "python",
         "args": ["-m", "yfin_mcp"]
       }
     }
   }
   ```

   **For development/source installation**:
   - macOS:
     ```json
     {
       "mcpServers": {
         "yfinance": {
           "command": "uv",
           "args": [
             "--directory",
             "/ABSOLUTE/PATH/TO/yahoo-finance-mcp",
             "run",
             "server.py"
           ]
         }
       }
     }
     ```
   - Windows:
     ```json
     {
       "mcpServers": {
         "yfinance": {
           "command": "uv",
           "args": [
             "--directory",
             "C:\\ABSOLUTE\\PATH\\TO\\yahoo-finance-mcp",
             "run",
             "server.py"
           ]
         }
       }
     }
     ```

4. **Restart Claude for Desktop**

### Development Mode

For testing with MCP Inspector:

```bash
# From source
uv run yfin-mcp

# Or if installed via pip
python -m yfin_mcp
```

## Publishing to PyPI

To build and publish the package, use the provided scripts. You can optionally provide an argument to bump the version:

### Windows
```bash
# Just build and publish current version
publish_package.bat

# Bump version and then publish
publish_package.bat patch
publish_package.bat minor
publish_package.bat major
```

### macOS/Linux
```bash
chmod +x publish_package.sh

# Just build and publish current version
./publish_package.sh

# Bump version and then publish
./publish_package.sh patch
./publish_package.sh minor
./publish_package.sh major
```

> [!NOTE]
> The scripts will build the package into the `dist/` directory and then use `twine` to upload it. Ensure you have your PyPI credentials configured in `~/.pypirc` (or `%HOME%\.pypirc` on Windows) or set the `TWINE_PASSWORD` environment variable.


## Troubleshooting

### [ERROR] os error 32: Process cannot access the file
If you see this error when running `publish_package.bat` or `uv build`, it means the MCP server is still running and locking the executable.
1. **Close Claude Desktop** or any app using the yfinance MCP server.
2. Stop any running **MCP Inspector** instances.
3. If the error persists, manually kill the processes:
   ```bash
   taskkill /F /IM yfin-mcp.exe /T
   taskkill /F /IM python.exe /T
   ```

## License

MIT License

**Original Work**: Copyright (c) 2025 AlexYoung  
**Fork Enhancements**: Copyright (c) 2026 SKTelecom

This project maintains the MIT License from the original [yahoo-finance-mcp](https://github.com/Alex2Yang97/yahoo-finance-mcp) project. All enhancements (pagination, caching, export) are also released under MIT License.

See [LICENSE](LICENSE) file for full details.


TDQS

B3.4/5.0

Scored across 9 tools

Disambiguation5/5

All tools target distinct financial data categories: financial statements, historical prices, holder info, options, recommendations, stock actions, general info, and news. No overlap in functionality.

Naming Consistency5/5

All tools follow a consistent 'get_' prefix with underscore-separated descriptive nouns (e.g., get_financial_statement, get_historical_stock_prices). The pattern is uniform and predictable.

Tool Count5/5

With 9 tools, the server covers the essential Yahoo Finance data domains without being bloated or sparse. Each tool serves a specific purpose, and the count is well-scoped for the server's purpose.

Completeness4/5

The set covers financial statements, historical prices, holders, options, recommendations, actions, general info, and news. Minor gaps like earnings call transcripts or more granular estimates exist, but the core data surface is nearly complete.