MATLAB MCP Server (matlab-mcp-python)
MCP serverCommunity
A Python MATLAB server with an engine pool for several users, async long jobs, and figures returned as interactive Plotly charts. By HanSur94.
4 starsUpdated 4 months agoGitHub data from 5 Nov 2026
At a glance
- Type
- MCP server
- Works with
- MATLAB
- How it connects
- Drives a running copy of the tool
- Access
- Can modify models
- Tool versions
- MATLAB R2022b or later with the MATLAB Engine API for Python
- Operating systems
- Windows, macOS, Linux
- Needs a MATLAB licence
- Yes
- Transport
- Local (stdio), Remote (HTTP)
- AI clients documented
- Claude Desktop, Claude Code, Cursor
- Licence
- MIT
- Language
- Python
What it can do
- Runs MATLAB code synchronously for short commands and switches long jobs to async with progress reporting.
- Scales a pool of MATLAB engines and isolates per-user workspaces when served over HTTP.
- Browses installed toolboxes, functions and help text, and runs checkcode before execution.
- Converts MATLAB figures to Plotly JSON with a static PNG fallback, and exposes your own .m functions as tools through a YAML file.
- Has a configurable blocklist of MATLAB functions the agent may not call.
Limitations
- The README lists Windows and macOS as supported platforms; Linux is covered only by the install commands.
- The matlabengine package isn't a dependency; you install the version that matches your MATLAB release yourself (the Pixi route has a task for this).
Install
pip install matlab-mcp-pythonSource: github.com/HanSur94/matlab-mcp-server-python#install-the-server
claude mcp add matlab -- matlab-mcpSource: github.com/HanSur94/matlab-mcp-server-python#connect-to-claude-code
matlab-mcp --transport streamablehttpMCP servers and skills run code on your computer. Review the source before installing.
AI-generated models, drawings and results must be checked by a qualified engineer before use.