FinanceToolkit

To build the MCP bundle from a local repository for testing purposes, follow these steps:

1. Clone the repository: git pull https://github.com/JerBouma/FinanceToolkit
2. Install dependencies: uv sync
3. Run the build script: bash financetoolkit/mcp_server/mcpb/build-mcpb.sh
4. The resulting bundle will be located in the dist/ directory.

Note: By default, the bundle uses the published PyPI package. To include uncommitted local changes in your bundle, use the --local flag. This produces dist/financetoolkit-local.mcpb, which points to your local checkout via uv's [tool.uv.sources].

To install the MCP server locally, you can use the automated setup wizard which locates your client's configuration file and writes the necessary MCP entry and API key automatically.

Alternatively, you can manually add the server configuration to your client's MCP config file (e.g., claude_desktop_config.json, .cursor/mcp.json, or .vscode/mcp.json).

uvx --from "financetoolkit[mcp]" financetoolkit-mcp-setup

You can query over 200 financial metrics from MCP-compatible clients (like Claude, Copilot, Cursor, or Windsurf) without writing code.

Local Setup

Use uvx to set up your client configuration and API key automatically:

uvx --from "financetoolkit[mcp]" financetoolkit-mcp-setup

Hosted Setup

Connect to the hosted server: https://financetoolkit.jeroenbouma.com/mcp (OAuth handles authentication on first use).

To begin using the toolkit, initialize the Toolkit class with a list of ticker symbols, your API key, and a starting date. This instance serves as the base for accessing various financial modules like ratios, technicals, and risk metrics.

from financetoolkit import Toolkit

# Initialize the Toolkit for Apple and Microsoft
companies = Toolkit(["AAPL", "MSFT"], api_key=API_KEY, start_date="2017-12-31")

You can install the Finance Toolkit MCP Server into compatible AI clients (like Claude Desktop or Claude Code) using the .mcpb bundle. This method requires no manual Python configuration.

1. Download the latest financetoolkit.mcpb from the releases page.
2. Double-click the downloaded file to open Claude Desktop.
3. Click "Install" when prompted by the Claude Desktop prompt.
4. Confirm the installation by clicking "Install" again.
5. Provide your Financial Modeling Prep API key (obtain one here) and click "Save".
6. Toggle the bundle switch from "Disabled" to "Enabled".
7. Restart Claude Desktop to begin using the Finance Toolkit MCP Server in your conversations.

Install the financetoolkit package using pip. It is recommended to use the -U flag to ensure you have the latest version.

pip install financetoolkit -U

You can connect directly to the hosted Finance Toolkit MCP server without local installation. On the first connection, you will be prompted via an OAuth consent page to provide your Financial Modeling Prep (FMP) API key.

Connection URLs and Methods:

• Claude Desktop / Claude.ai: Customize → Connectors → Add custom connector → paste https://financetoolkit.jeroenbouma.com/mcp
• Claude Code: Use the CLI command provided below.
• VS Code: Command Palette → MCP: Add Server → HTTP → paste the URL.
• Cursor: Settings → Features → MCP Servers → Add new → http → paste the URL.
• Windsurf: Settings → MCP Servers → Add Server → Remote/HTTP → paste the URL.

claude mcp add --transport http finance-toolkit https://financetoolkit.jeroenbouma.com/mcp

The Finance Toolkit provides three powerful capabilities across nearly all get_* and collect_* functions to transform raw data into time series or standardized scores:

• rolling=<n> and trailing=<n>: Use rolling for a sliding window or trailing for a trailing sum/average (e.g., trailing=4 for a trailing 4-quarter sum) to turn snapshots into time series.
• growth=True and lag: Pass growth=True to return period-over-period growth. Use lag (an int or list of ints, default 1) to specify the comparison period (e.g., lag=4 for year-over-year growth on quarterly data). Combine with trailing (e.g., trailing=4, growth=True) to calculate Trailing Twelve Months (TTM) growth.
• standardize=True: Converts raw values into Z-Scores (standard deviations from their historical mean/std). This is useful for ranking or comparing metrics on different scales.

If you prefer manual configuration, add the following block to your MCP configuration file. Ensure you replace YOUR_API_KEY_HERE with your actual Financial Modeling Prep API key.

{
  "mcpServers": {
    "finance-toolkit": {
      "command": "uvx",
      "args": ["--from", "financetoolkit[mcp]", "financetoolkit-mcp"],
      "env": { "FINANCIAL_MODELING_PREP_API_KEY": "YOUR_API_KEY_HERE" }
    }
  }
}

If you are using a free FinancialModelingPrep plan or encounter API limits, the toolkit can fall back to Yahoo Finance to retrieve missing data. This requires the yfinance library to be installed in your environment.

To enable this functionality, run:

pip install yfinance

Use the setup() function (exposed via the financetoolkit-mcp-setup command) to configure the Finance Toolkit for various Model Context Protocol (MCP) clients.

Interactive Mode

Running setup() without arguments launches an interactive wizard that allows you to select one or more clients (Claude Desktop, Claude Code, VS Code, Cursor, Gemini, Windsurf) and manage your API key.

Non-interactive Mode

To configure a specific client without the interactive menu, use the --client flag. This is useful for automated environments or scripts.

Arguments

• --client {claude-desktop,claude-code,vscode,cursor,gemini,windsurf}: Configure a specific client non-interactively.
• --overwrite: Silently overwrite an existing configuration without prompting.