Math MCP Server

Overview

This is an MCP (Model Context Protocol) server that can do basic arithmetic in 64 bit precision, along with matrix multiplication.

Features

• Addition

• Subtraction

• Multiplication

• Division

• Matrix multiplication

Installation

Prerequisities

Ensure you have the following installed:

• Python 3.13+
• uv

Installing UV

See installation guide for all options.

Unix/MacOS

curl -LsSf https://astral.sh/uv/install.sh | sh

or

brew install uv

Windows

winget install --id=astral-sh.uv  -e

Clone the Repository

git clone https://github.com/avanishd-3/math-mcp.git
cd math-mcp-server
uv sync

Integration with Clients

Claude Code

fastmcp install claude-code src/math_server.py:math_mcp

Claude Desktop

fastmcp install claude-desktop src/math_server.py:math_mcp

Cursor

fastmcp install cursor src/math_server.py:math_mcp

VS Code

Add the following .vscode/mcp.json and use your actual path.

{
  "servers": {
    "Math MCP Server": {
      "command": "uv",
      "args": [
        "run",
        "--with",
        "fastmcp",
        "--with",
        "numpy",
        "fastmcp",
        "run",
        "/absolute/path/Desktop/to/math-mcp-server/src/math_server.py:math_mcp"
      ]
    }
  },
}

Contributing

1. Fork the repository

2. Create a new branch:

git checkout -b add-feature

3. Make changes and commit (remember to add unit tests in test/ directory)

git commit -m "Added a new feature"

4. Push to your fork git push origin add-feature

5. Open a pull request.

Project Structure

/
├── src
│   └── math_server.py
├── tests
│   ├── test_arithmetic.py
│   │   └── astro.svg
│   ├── test_linear_algebra.py
├── pytest.ini
├── pyproject.toml
└── uv.lock

Architecture

This MCP server uses Fast MCP 2.0, which provides many more features than Fast MCP 1.0, which is what the official Python SDK for MCP uses.

Also, the unit tests are written with Pytest, which is what Fast MCP 2.0 recommends.

Lastly, if you don't know, uv is a much faster version of pip that also provides a lockfile for project dependencies (this will be familiar if you've used npm or cargo before). The MCP Python SDK itself uses uv, and I use it for all new Python projects, because it's 10-100x faster than pip, and the lockfile makes dependency version management much simpler.