math -learning

The Math MCP Learning Server exposes a suite of 17 tools that provide persistent workspace management, advanced mathematical computations, matrix operations, and visualization capabilities. It is built atop the FastMCP framework and uses the Model Context Protocol Python SDK, enabling clients to connect via a wide range of MCP clients. Features include a persistent workspace for saving calculations, history tracking, and a library of mathematical functions and constants. Tools are categorized into Workspace, Math, Matrix, and Visualization, allowing you to save or load calculations, safely evaluate expressions, perform statistical analysis, handle unit conversions, manipulate matrices, and generate plots or charts for data analysis. Cloud hosting is supported, making it suitable for hosted tutoring, teaching assistants, or AI agents that rely on reliable math tooling.

To use the server, connect with your MCP client of choice (Claude Desktop, Claude Code, Goose, OpenCode, Kiro, Gemini CLI, or any MCP-compatible client) and point to the math-mcp-learning-server instance. Local or cloud deployment is supported; the provided commands enable you to run the server locally via uvx or connect to the hosted cloud endpoint if you prefer no installation. Once connected, you can perform calculations, save results to the workspace, load saved variables, plot functions, generate histograms, and run matrix operations directly through the MCP interface.

Common workflows include evaluating expressions with the calculate tool, performing matrix operations such as matrix_multiply or matrix_inverse, visualizing data with plot_function or create_histogram, and persisting important results with save_calculation for later reference. Prompts like math_tutor and formula_explainer can guide users with structured tutoring prompts and step-by-step formula explanations, respectively.

Prerequisites

• Python 3.8+ and pip installed on your system
• An MCP client installed (see Quick Start in the README) or cloud access to the hosted server

Option A: Install and run locally using uvx (recommended per the project README)

1. Install uvx (if not already installed):

# Using pipx (recommended)
pipx install uvx

# Or, if you prefer a direct pip install (less isolated):
pip install uvx

2. Run the server locally:

uvx math-mcp-learning-server

3. Optional: install with additional features (scientific calculations, plotting, etc.):

# All features
uvx --from 'math-mcp-learning-server[scientific,plotting]' math-mcp-learning-server

# Scientific features only
uvx --from 'math-mcp-learning-server[scientific]' math-mcp-learning-server

# Plotting features only
uvx --from 'math-mcp-learning-server[plotting]' math-mcp-learning-server

Option B: Automatic/Multi-feature installation via uvx (as shown in the README)

uvx --from 'math-mcp-learning-server[scientific,plotting]' math-mcp-learning-server

Notes

• The server runs as a local process and exposes an MCP endpoint that clients connect to using the standard MCP transport (HTTP by default for cloud deployments, or local transport for development).
• If you are deploying to the cloud, you can point your MCP client to the hosted endpoint and avoid local installation.

Tips and common issues:

• Ensure your MCP client is compatible with the Model Context Protocol Python SDK and supports the 17 available tools.
• When using the local installation, the workspace directory is persistent by design; ensure you have sufficient disk space for history and saved computations.
• If you encounter permission or path issues on Linux/macOS, ensure Python and the uvx binary are on your PATH.
• For advanced usage, you can enable optional features like scientific computations and plotting by installing the corresponding extras (shown in the installation section).
• Environment variables can be used to customize the server behavior in some deployment scenarios (e.g., endpoints, timeouts) depending on your hosting environment.