Local MCP Server

Overview

This guide covers the steps required to run the Kensho LLM-Ready API MCP (Model Context Protocol) (opens in a new tab) server locally. For managed hosting options, please refer to our Remote MCP Server integrations.

The guide includes information on ensuring an appropriate Python version is installed, how to install the LLM-Ready API client package, and finally how to connect to the MCP server using Claude Desktop (opens in a new tab).

Python Installation

For users who do not already have Python>=3.10 installed, please follow instructions here (opens in a new tab) to get set up.

LLM-Ready API Installation

Installation instructions are provided in the kFinance (opens in a new tab) GitHub repo, but for clarity they are reproduced here. Simply run:

pip install kensho-kfinance

Then confirm in the output that the installation completed successfully.

Adding the local MCP server to Claude Desktop

Below are the basic steps required to add the LLM-Ready API MCP server to Claude Desktop’s available tools. More detailed information can be found on the official MCP documentation (opens in a new tab).

Open Claude Desktop.
Navigate to the Settings panel.
In the “Developer” tab, click “Edit Config”.
Open claude_desktop_config.json in a text editor like Notepad or TextEdit.
Paste the following settings into the config, replacing <path_to_python> below with your appropriate path which can be found by running which python.

{
  "mcpServers": {
    "kfinance": {
      "command": "<path_to_python>",
        "args": [
          "-m",
          "kfinance.mcp",
          "--stdio"
        ]
     }
   }
}

Exit and restart Claude Desktop.