Labsco
MCP SERVER

Math MCP Learning

by clouatre-labs ↗

Evaluate expressions, run matrix algebra and draw charts from the conversation — and save a result under a name so the next session can load it back.

Physical Sciences & Scientific Computing✓ Verified
Summary
Arithmetic the model does not have to do in its head, and a chart at the end of it.

Routing calculation to a tool removes a familiar failure: the numbers come from an evaluator rather than from a language model's arithmetic, and the workspace keeps a named result available across sessions rather than in the transcript. Install the scientific and plotting extras if you want the matrix and chart tools at all — without NumPy and matplotlib those calls raise. One tool to read carefully: the financial line chart generates synthetic price movement for teaching, not real market data.

What it is

A maths and plotting toolkit with a workspace that persists. Expressions, statistics, unit conversion and linear algebra return numbers; the plot tools return images; saved calculations survive a restart under the name you gave them.

What you get
  • Expression evaluation over +, -, *, /, ** and parentheses with sin, cos, tan, log, sqrt, abs and pow
  • Descriptive statistics over a list — mean, median, mode, standard deviation, variance
  • Compound interest from principal, rate, time and compounding frequency, and unit conversion across length, weight and temperature
  • Linear algebra: multiply, transpose, determinant, inverse and eigenvalues
  • Plots as images: a function over an x-range, histograms, line charts, scatter plots and box plots comparing several groups, each with labels, colour and grid control
  • A named calculation saved with its expression and result, and loaded back in a later session
  • Readable resources alongside the tools — the workspace summary, calculation history, the function reference, mathematical constants and the active variables
  • Two prompts: a tutoring prompt with a difficulty setting, and a formula explainer
Requirements

Nothing to authorize. Hosted over Streamable HTTP at https://math-mcp.fastmcp.app/mcp for zero install, or locally over stdio as uvx math-mcp-learning-server[scientific,plotting] — those two extras are what bring in the dependencies the optional tools need, since the matrix tools require NumPy and every plot tool requires matplotlib, and they raise rather than degrade without them. Apache-2.0.

Setup effort

One command — uvx math-mcp-learning-server