Summary

It runs the code you give it, on your machine.

This is a thin execution bridge rather than an animation designer: the quality of the result is entirely the quality of the Manim script the model writes, and the server's job is to render it and hand back the file. Because it executes Python locally with your Manim install, treat it like any other code-running tool — the sandbox is your own machine.

What it is

An MCP server that executes Manim scripts and returns the generated animation. The model writes the Manim Python, the server runs it and saves the output into a visible media folder, and temporary files can be cleaned up afterwards.

What you get

Execution of Manim Python scripts and the rendered video as the result
Animation output written to a visible media folder rather than a temp path you have to hunt for
A cleanup step for the temporary files an execution leaves behind
Configuration through environment variables, so the Manim binary can live anywhere

Requirements

Python 3.8+, the community edition of Manim (`pip install manim`) and the `mcp` package. Point `MANIM_EXECUTABLE` at your Manim binary and launch the server as `src/manim_server.py` with an absolute Python path — the README shows how to find both on Windows, macOS and Linux. The project is `manim-mcp-server` (0.1.0 in pyproject); installation is a clone. MIT licensed.

Setup effort

One command — pip install manim