Content
# MATLAB MCP Server
<p align="center">
English •
<a href="l10n/README.es.md">Español</a> •
<a href="l10n/README.ja.md">日本語</a> •
<a href="l10n/README.ko.md">한국어</a> •
<a href="l10n/README.zh-cn.md">简体中文</a>
</p>
Run MATLAB® using AI applications with the official MATLAB MCP Server from MathWorks®. The MATLAB MCP Server allows your AI applications to:
- Start and quit MATLAB.
- Write and run MATLAB code.
- Assess your MATLAB code for style and correctness.
To assist your agent in using MATLAB and Simulink, you can use skills from [MATLAB Agentic Toolkit (GitHub)](https://github.com/matlab/matlab-agentic-toolkit) and [Simulink Agentic Toolkit (GitHub)](https://github.com/matlab/simulink-agentic-toolkit), which can also install this MCP server for you.
## Table of Contents
- [Setup](#setup)
- [Claude Code](#claude-code)
- [Claude Desktop](#claude-desktop)
- [GitHub Copilot in Visual Studio Code](#github-copilot-in-visual-studio-code)
- [Codex](#codex)
- [Arguments](#arguments)
- [Tools](#tools)
- [Resources](#resources)
- [Data Collection](#data-collection)
- [Security Considerations](#security-considerations)
- [Licensing and Usage](#licensing-and-usage)
- [Contact Support](#contact-support)
## Setup
1. Install [MATLAB (MathWorks)](https://www.mathworks.com/help/install/ug/install-products-with-internet-connection.html) R2021a or later and add it to the system PATH. The MATLAB MCP Server supports MATLAB releases from the past five years.
1. The server supports any AI application that uses Model Context Protocol. To set up the MATLAB MCP Server for Claude Desktop, skip to the instructions for [Claude Desktop](#claude-desktop). To set up the server for other applications, follow these instructions:
- For Windows or Linux, [**Download the Latest Release**](https://github.com/matlab/matlab-mcp-server/releases/latest). (Alternatively, you can **build from source**: install [Go](https://go.dev/doc/install) and build the binary using `go install github.com/matlab/matlab-mcp-server/cmd/matlab-mcp-server@latest`).
- For macOS, first download the latest release by running the following command in your terminal:
- For Apple silicon processors, run:
```sh
curl -L -o ~/Downloads/matlab-mcp-server https://github.com/matlab/matlab-mcp-server/releases/latest/download/matlab-mcp-server-macos-arm64
```
- For Intel processors, run:
```sh
curl -L -o ~/Downloads/matlab-mcp-server https://github.com/matlab/matlab-mcp-server/releases/latest/download/matlab-mcp-server-macos-x64
```
Then grant executable permissions to the downloaded binary so you can run the MATLAB MCP Server:
```sh
chmod +x ~/Downloads/matlab-mcp-server
```
1. Add the MATLAB MCP Server to your AI application. You can find instructions for adding MCP servers in the documentation of your AI application. For example instructions on using Claude Code®, Claude Desktop®, and GitHub Copilot in Visual Studio® Code, see below. Note that you can customize the server by specifying optional [Arguments](#arguments).
### Claude Code
In your terminal, run the following, remembering to insert the full path to the server binary you acquired in the setup:
```sh
claude mcp add --transport stdio matlab -- /fullpath/to/matlab-mcp-server-binary
```
You can customize the server by specifying optional [Arguments](#arguments). Note the `--` separator between Claude Code's options and the server arguments:
```sh
claude mcp add --transport stdio matlab -- /fullpath/to/matlab-mcp-server-binary --initial-working-folder=/home/username/myproject
```
For details on adding MCP servers in Claude Code, see [Add a local stdio server (Claude Code)](https://docs.claude.com/en/docs/claude-code/mcp#option-3%3A-add-a-local-stdio-server). To remove the server later, run:
```sh
claude mcp remove matlab
```
### Claude Desktop
You install the MATLAB MCP Server in Claude Desktop using the MATLAB MCP Server bundle.
1. Install the Filesystem extension in Claude Desktop to allow Claude to read and write files on your system. In Claude Desktop, click **Settings > Extensions > Browse extensions**. Search for the Filesystem extension developed by Anthropic and click **Install**. Specify the folders you want to allow the MCP server to access, then toggle the **Disabled** button to **Enable** the Filesystem extension.
2. Download the MATLAB MCP Server bundle `matlab-mcp-server.mcpb` from the [Latest Release](https://github.com/matlab/matlab-mcp-server/releases/latest) page.
3. To install the MATLAB MCP Server bundle as a desktop extension, double click the downloaded `matlab-mcp-server.mcpb` file and click **Install** in Claude Desktop. (Alternatively, navigate in Claude to **File menu > Settings > Extensions > Advanced Settings > Install Extension** and select the `matlab-mcp-server.mcpb` file. Click **Install**).
To customize the behaviour of the MATLAB MCP Server, navigate to **Settings > Extensions > Configure**, where you can modify the server's [Arguments](#arguments).
### GitHub Copilot in Visual Studio Code
In your VS Code workspace, create a file named `.vscode/mcp.json`. Insert the following JSON, remembering to specify the full path to the server binary you acquired in the setup, as well as any [Arguments](#arguments). Then save the file. (Note that on Windows, your paths require extra slashes as escape characters).
```json
{
"servers": {
"matlab": {
"type": "stdio",
"command": "C:\\fullpath\\to\\matlab-mcp-server-windows-x64.exe",
"args": []
}
}
}
```
For more information about using MCP servers in VS Code, see [Add and Manage MCP servers in VS Code (VS Code)](https://code.visualstudio.com/docs/copilot/customization/mcp-servers#_configure-the-mcpjson-file).
### Codex
In your terminal, run the following, remembering to insert the full path to the server binary you acquired in the setup:
```sh
codex mcp add matlab -- /fullpath/to/matlab-mcp-server-binary
```
You can customize the server by specifying optional [Arguments](#arguments). For example:
```sh
codex mcp add matlab -- /fullpath/to/matlab-mcp-server-binary --initial-working-folder=/home/username/myproject
```
> [!NOTE]
> **Windows users:** Codex does not forward the `WINDIR` environment variable, which you need to run MATLAB and Simulink (see #32).
>
> To add the variable, open your Codex config file at `C:\Users\<username>\.codex\config.toml`, and add `env_vars = ["WINDIR"]` to the `[mcp_servers.matlab]` section:
>
> ```toml
> [mcp_servers.matlab]
> command = 'C:\fullpath\to\matlab-mcp-server-windows-x64.exe'
> args = []
> env_vars = ["WINDIR"]
> ```
## Arguments
Customize the behavior of the server by specifying arguments in one of these ways:
- Insert the arguments in the configuration settings of your AI application (usually a `.json` file).
- Enter the arguments as command-line interface (CLI) flags when you start the server.
- Use environment variables, specified either in your CLI or application's configuration settings. To derive the environment variable name from a CLI flag, add the prefix `MW_MCP_SERVER_`, convert to uppercase, and replace hyphens (`-`) with underscores (`_`). For example, the argument `--matlab-root` becomes the environment variable `MW_MCP_SERVER_MATLAB_ROOT`. CLI flags take precedence over environment variables, if you use both.
| Argument | Description | Example |
| ------------- | ------------- | ------------- |
| help | Displays help information for all arguments. | `--help` |
| version | Displays the version of the MATLAB MCP Server. | `--version` |
| matlab-root | Full path specifying which MATLAB to start. Do not include `/bin` in the path. By default, the server tries to find the first MATLAB on the system PATH. | Windows: `--matlab-root=C:\\Program Files\\MATLAB\\R2026a` <br><br> Linux/macOS: `--matlab-root=/home/usr/MATLAB/R2026a`<br><br>As an environment variable: `MW_MCP_SERVER_MATLAB_ROOT=/home/usr/MATLAB/R2026a` |
| initialize-matlab-on-startup | To initialize MATLAB as soon as you start the server, set this argument to `true`. By default, MATLAB only starts when the first tool is called. | `--initialize-matlab-on-startup=true` |
| initial-working-folder | Specify the folder where MATLAB starts. If you do not specify a value, MATLAB starts at the path of your AI application's first [Root (MCP)](https://modelcontextprotocol.io/specification/latest/client/roots). If you have not defined a root, MATLAB starts in these locations: <br> <ul><li>Linux: `/home/username` </li><li> Windows: `C:\Users\username\Documents`</li><li>Mac: `/Users/username/Documents`</li></ul> | Windows: `--initial-working-folder=C:\\Users\\username\\MyProject` <br><br> Linux/macOS: `--initial-working-folder=/Users/username/MyProject` |
| matlab-display-mode | Specify whether to show the MATLAB desktop. Use `desktop` mode (default) to show the MATLAB desktop. Use `nodesktop` mode to use MATLAB only from your AI application, without the MATLAB desktop. Note that in `nodesktop` mode, commands requiring a graphical interface (such as `edit`, `open`, `open_system`, `uifigure`, and `appdesigner`) will still open MATLAB windows on your desktop. | `--matlab-display-mode=nodesktop` |
| matlab-session-mode | Specify whether the MCP server starts a new MATLAB or connects to an existing MATLAB session (supported for MATLAB R2023a onwards). The default is **`auto`** mode.<br><br> **`new` mode:** The MCP server starts a new MATLAB session. <br><br>**`auto` mode (default):** The server tries to connect to an existing MATLAB session, which you must have configured for `existing` mode using the instructions below. If the server is unable to find an existing MATLAB session, it starts a new one. <br><br>**`existing` mode:** The server tries to connect to an existing MATLAB session. You must have configured your MATLAB session beforehand to use this mode, with these steps:<br><br><ol><li>If you are using `existing` mode for the first time, run `./matlab-mcp-server --setup-matlab`.<br><br>This command installs an add-on named MATLAB MCP Server Toolbox in MATLAB. You can customize the command with other arguments from this table. For example, to specify which MATLAB to use to install the toolbox, you can use `./matlab-mcp-server --setup-matlab --matlab-root=/home/usr/MATLAB/R2026a`.<br><br>For Claude Desktop, you must download the MATLAB MCP Server binary using the instructions in [Setup](#setup) before you run `./matlab-mcp-server --setup-matlab`.<br><br></li><li>In the command window of a running MATLAB session, run `shareMATLABSession()`. The MCP server will connect to this MATLAB when you start the server with `--matlab-session-mode=existing` or `--matlab-session-mode=auto`. If you are running multiple MATLAB sessions, the server connects to the MATLAB session where you most recently ran the command `shareMATLABSession()`.<br><br>As an alternative to running `shareMATLABSession()` manually, you can add the command to your MATLAB [Startup Script (MathWorks)](https://www.mathworks.com/help/matlab/ref/startup.html).</li></ol> | `--matlab-session-mode=existing` |
| extension-file | To use custom MCP tools, provide a path to a JSON file that defines your tools. You can also use multiple extension files. For details on using custom tools, see [Use Custom Tools with the MATLAB MCP Server](guides/custom-tools.md). | <br><br>Windows: `--extension-file=C:\\Users\\name\\my-tools.json` <br><br> Linux/macOS: `--extension-file=/path/to/my-tools.json` <br><br> **Using multiple extension files:**<br><br>Windows:`--extension-file=C:\\path\\to\\tools-1.json --extension-file=C:\\path\\to\\tools-2.json`<br><br>Linux/macOS:`--extension-file=/path/to/tools1.json --extension-file=/path/to/tools2.json` <br><br> **Using environment variables:** <br><br> Windows: `MW_MCP_SERVER_EXTENSION_FILE=C:\Users\name\tools1.json;C:\Users\name\tools2.json` <br><br> Linux/macOS: `MW_MCP_SERVER_EXTENSION_FILE=/path/to/tools1.json:/path/to/tools2.json` |
| log-folder | Specify the folder where the MCP server stores log files. If not specified, the server uses the default temporary folder of your operating system. | Windows: `--log-folder=C:\\Users\\name\\AppData\\Local\\Temp` <br><br> Linux/macOS: `--log-folder=/tmp/my-logs` |
| log-level | The log levels of the MCP server. Valid values, in order of decreasing verbosity, are `debug`, `info`, `warn`, and `error`. The default is `info`. | `--log-level=debug` |
| disable-telemetry | To disable anonymized data collection, set this argument to `true`. For details, see [Data Collection](#data-collection). | `--disable-telemetry=true` |
## Tools
1. `detect_matlab_toolboxes`
- Returns information about installed MATLAB and toolboxes, including version numbers.
1. `check_matlab_code`
- Performs static code analysis on a MATLAB script. Returns warnings about coding style, potential errors, deprecated functions, performance issues, and best practice violations. This is a non-destructive, read-only operation that helps identify code quality issues without executing the script.
- Inputs:
- `script_path` (string): Absolute path to the MATLAB script file to analyze. Must be a valid `.m` file. The file is not modified during analysis. Example: `C:\Users\username\matlab\myFunction.m` or `/home/user/scripts/analysis.m`.
1. `evaluate_matlab_code`
- Evaluates a string of MATLAB code and returns the output.
- Inputs:
- `code` (string): MATLAB code to evaluate.
- `project_path` (string): Absolute path to your project directory. MATLAB sets this directory as the current working folder. Example: `C:\Users\username\matlab-project` or `/home/user/research`.
1. `run_matlab_file`
- Executes a MATLAB script and returns the output. The script must be a valid `.m file`.
- Inputs:
- `script_path` (string): Absolute path to the MATLAB script file to execute. Must be a valid `.m` file. Example: `C:\Users\username\projects\analysis.m` or `/home/user/matlab/simulation.m`.
1. `run_matlab_test_file`
- Executes a MATLAB test script and returns comprehensive test results. Designed specifically for MATLAB unit test files that follow MATLAB testing framework conventions.
- Inputs:
- `script_path` (string): Absolute path to the MATLAB test script file. Must be a valid `.m` file containing MATLAB unit tests. Example: `C:\Users\username\tests\testMyFunction.m` or `/home/user/matlab/tests/test_analysis.m`.
## Resources
The MCP server provides [Resources (MCP)](https://modelcontextprotocol.io/specification/latest/server/resources) to help your AI application write MATLAB code. To see instructions for using this resource, refer to the documentation of your AI application that explains how to use resources.
1. `matlab_coding_guidelines`
- Provides comprehensive MATLAB coding standards for improving code readability, maintainability, and collaboration. The guidelines encompass naming conventions, formatting, commenting, performance optimization, and error handling.
- URI: `guidelines://coding`
- MIME Type: `text/markdown`
- Source: [MATLAB Coding Standards (GitHub)](https://github.com/matlab/rules/blob/main/matlab-coding-standards.md)
1. `plain_text_live_code_guidelines`
- Provides rules and guidelines for generating live scripts using the plain text Live Code `.m` file format, suitable for version control and AI-assisted development. Note that to run plain text live scripts you need MATLAB R2025a or newer. For details, see [Live Code File Format (MathWorks)](https://www.mathworks.com/help/matlab/matlab_prog/plain-text-file-format-for-live-scripts.html).
- URI: `guidelines://plain-text-live-code`
- MIME Type: `text/markdown`
- Source: [Plain Text Live Code Generation (GitHub)](https://github.com/matlab/rules/blob/main/live-script-generation.md)
## Data Collection
The MATLAB MCP Server may collect fully anonymized information about your usage of the server and send it to MathWorks. This data collection helps MathWorks improve products and is on by default. To opt out of data collection, set the argument `--disable-telemetry` to `true`.
## Security Considerations
When using the MATLAB MCP Server, you should thoroughly review and validate all tool calls before you run them. Always keep a human in the loop for important actions and only proceed once you are confident the call will do exactly what you expect. For more information, see [User Interaction Model (MCP)](https://modelcontextprotocol.io/specification/latest/server/tools#user-interaction-model) and [Security Considerations (MCP)](https://modelcontextprotocol.io/specification/latest/server/tools#security-considerations).
## Licensing and Usage
The license is available in the [LICENSE.md](LICENSE.md) file in this GitHub repository.
MCP servers are only permitted to be used with MATLAB in accordance with the MathWorks Software License Agreement, and must not be shared by multiple users. Contact MathWorks if you need to support shared or centralized server use.
## Contact Support
MathWorks encourages you to use this repository and provide feedback. To request technical support or submit an enhancement request, [create a GitHub issue](https://github.com/matlab/matlab-mcp-server/issues) or contact [MathWorks Technical Support](https://www.mathworks.com/support/contact_us.html).
---
Copyright 2025-2026 The MathWorks, Inc.
---
