Market Index

You can search and compare long histories of market indicators like S&P 500, NASDAQ-100, and Nikkei 225 using natural language. This MCP server is fast, locally cached, and integrates with Claude Desktop so you can ask in plain language and get precise historical data and summaries.

How to use

Ask natural language questions to Claude Desktop about market indices. For example, you can request the lowest and highest prices for a specific period, compare two indices over a year, identify top drawdown or best days in a given year, or compute annual returns. The MCP provides data such as price ranges, daily prices, yearly summaries, and performance comparisons across supported symbols like S&P 500, NASDAQ 100, Dow Jones, Nikkei 225, FTSE 100, and VIX.

Typical uses include: getting a range of prices for a date range, comparing multiple indices in a single query, ranking worst or best days in a year, and calculating annual returns for a specific index. You can mix questions to build multi-part requests that Claude Desktop resolves using the local MCP cache for fast results.

How to install

Prerequisites you need before installing the MCP server:

• Python 3.10 or higher

• uv or pip

• Claude Desktop

Follow these steps to set up and run the MCP server locally.

Step-by-step setup with Claude Desktop

{
  "mcpServers": {
    "market_index": {
      "command": "/opt/homebrew/bin/uvx",
      "args": [
        "--from", "git+https://github.com/tamappe/market-index-mcp.git",
        "market-index-mcp"
      ]
    }
  }
}

Initial Claude Desktop integration and data download

1. Confirm the uvx path on your system.

2. Edit the Claude Desktop configuration so it can launch the MCP server automatically.

3. Restart Claude Desktop to trigger the initial data download. The process downloads ~10 MB and completes in about 30 seconds.

Manual installation for development

If you prefer to run the MCP server directly from source, you can install and run it in development mode.

git clone https://github.com/tamappe/market-index-mcp.git
cd market-index-mcp
pip install -e .

# Run the MCP server
market-index-mcp

Usage examples you can try in Claude Desktop

• “Tell me the lowest and highest price for S&P 500 in March 2020.”

• “Compare the 2020 performance of S&P 500 and NASDAQ 100.”

• “What were the top 5 worst Dow days in 2008?”

• “What is the 2023 annual return for Nikkei 225?”

• “How high did the VIX reach during the COVID-19 shock period?”

Troubleshooting and tips

If the server does not start or you see a disconnection, check the logs for errors and verify the uvx path in the configuration.

If you encounter SSL certificate errors, update certifi for Python or install certificates as appropriate for your Python version.

If you see an Unexpected token error, clear the uvx cache and restart Claude Desktop.

If data download fails due to network issues, verify network connectivity and proxy settings if applicable, then retry.

To completely clear cache, remove the uvx cache and restart Claude Desktop.

Technical notes and architecture

The MCP stores data locally for fast querying and updates from a data source repository. The workflow is:

• Local SQLite cache stores data
• MCP Server serves queries via Claude Desktop
• Data is updated from a data source release when available
• First run downloads the market data package and caches it locally for quick responses.

Licensing

MIT license.

Available tools

list_symbols

Fetches the list of available market symbols that the MCP can query.

get_price_range

Returns the price range (high, low, average) for a specified period.

get_price_on_date

Gets price data for a specific date.

compare_symbols

Compares performance across multiple symbols.

get_worst_days

Returns the worst down days for a specified year.

get_best_days

Returns the best up days for a specified year.

get_yearly_summary

Provides yearly summary data such as year-to-date return.