Menu
Connect AI agents to MATLAB with MCP tools and curated skills for engineering and scientific workflows
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Basé sur la classification professionnelle SOC
Train and deploy Qwen-AgentWorld, a native language world model that simulates agentic environments across 7 domains (MCP, Search, Terminal, SWE, Android, Web, OS) for agent training and evaluation.
Build and orchestrate collaborative multi-agent teams using HiClaw's Manager-Workers architecture on Matrix with Kubernetes-native control.
Set up and manage collaborative multi-agent teams using HiClaw, a Kubernetes-native platform with Matrix rooms for human-in-the-loop AI coordination
Build and orchestrate collaborative multi-agent teams using HiClaw's Manager-Workers architecture with Matrix rooms, MCP servers, and Kubernetes-native deployment.
Deploy and orchestrate collaborative multi-agent teams using HiClaw's Manager-Workers architecture on Docker or Kubernetes with Matrix rooms for human oversight
Use Agent Apprenticeship to train AI agents through real-world tasks, reusable experience, and ecosystem learning signals
| name | matlab-agentic-toolkit |
| description | Connect AI agents to MATLAB with MCP tools and curated skills for engineering and scientific workflows |
| triggers | ["set up MATLAB for AI agents","configure MATLAB MCP server","install MATLAB agentic toolkit","connect agent to MATLAB","use MATLAB with AI coding assistant","run MATLAB code through agent","configure MATLAB skills for agent","troubleshoot MATLAB MCP connection"] |
Skill by ara.so — AI Agent Skills collection.
The MATLAB Agentic Toolkit enables AI coding agents to work with MATLAB by providing the Model Context Protocol (MCP) server and curated skills. This toolkit gives agents the knowledge and context to write idiomatic MATLAB code, run tests, diagnose errors, and follow best practices without hallucinating functions or missing features.
Clone the repository:
git clone https://github.com/matlab/matlab-agentic-toolkit.git
cd matlab-agentic-toolkit
Ask your AI agent to set up the toolkit:
Set up the MATLAB Agentic Toolkit
The agent will:
Start a new session after setup completes.
If you already have the MATLAB MCP Core Server installed:
claude plugin marketplace add "https://github.com/matlab/matlab-agentic-toolkit"
claude plugin install matlab-core@matlab-agentic-toolkit
This adds skills without modifying your existing MCP configuration.
For users who want both MATLAB and Simulink toolkits or need advanced configuration:
agenticToolkitInstaller.mltbx from the Simulink Agentic Toolkit releasessetupAgenticToolkit
This installer supports:
--matlab-session-mode=existing)After installation, your agent can use these MCP tools to interact with MATLAB:
Execute MATLAB code and return command window output:
Run this MATLAB code: x = linspace(0, 2*pi, 100); y = sin(x); plot(x, y);
The agent will use evaluate_matlab_code to execute and return results.
Execute a MATLAB script or function file:
Run the file analyze_data.m with the current data
Execute MATLAB tests via runtests with structured results:
Run tests in test_signal_processing.m and show the results
Perform static analysis using Code Analyzer:
Check this MATLAB code for issues: function out = calc(x) out = x^2 end
List installed MATLAB version and available toolboxes:
What version of MATLAB is running? List the installed toolboxes.
The server provides these resources for context:
Skills provide domain-specific expertise to your agent. The toolkit includes:
MATLAB Core — Foundational MATLAB programming, syntax, and workflows
MATLAB Data Import and Analysis — Data loading, preprocessing, and analysis
MATLAB Software Development — Professional development practices
MATLAB App Building — Creating interactive applications
Automotive — Automotive engineering workflows
Computational Biology — Bioinformatics and systems biology
Image Processing and Computer Vision — Image analysis and vision algorithms
Robotics and Autonomous Systems — Robotics development
Signal Processing — Digital signal analysis
Wireless Communications — Communication system design
RF and Mixed Signal — RF circuit design and analysis
Test and Measurement — Instrument control and data acquisition
Reporting and Database Access — Report generation and database connectivity
Ask your agent to execute code directly:
Create a matrix A with random values and compute its eigenvalues
Agent uses evaluate_matlab_code:
A = rand(5, 5);
eigenvalues = eig(A);
disp(eigenvalues);
Write a function to calculate the moving average of a signal,
then create tests for it
Agent creates movingAverage.m:
function y = movingAverage(x, windowSize)
% MOVINGAVERAGE Compute moving average of input signal
% y = movingAverage(x, windowSize) returns the moving average
% of x using a window of size windowSize
arguments
x (:,1) double
windowSize (1,1) double {mustBePositive, mustBeInteger}
end
b = ones(1, windowSize) / windowSize;
y = filter(b, 1, x);
end
And test file test_movingAverage.m:
classdef test_movingAverage < matlab.unittest.TestCase
methods(Test)
function testBasicOperation(testCase)
x = [1 2 3 4 5]';
y = movingAverage(x, 3);
testCase.verifyEqual(length(y), length(x));
end
function testConstantSignal(testCase)
x = ones(10, 1) * 5;
y = movingAverage(x, 3);
% After initial transient, output should equal input
testCase.verifyEqual(y(3:end), x(3:end), 'AbsTol', 1e-10);
end
end
end
Then runs tests: run_matlab_test_file('test_movingAverage.m')
Load data.csv, clean it, and create a visualization showing trends over time
Agent creates complete workflow:
% Import data
data = readtable('data.csv');
% Clean data - remove rows with missing values
cleanData = rmmissing(data);
% Convert to timetable if date column exists
if ismember('Date', cleanData.Properties.VariableNames)
cleanData.Date = datetime(cleanData.Date);
tt = table2timetable(cleanData);
% Plot time series
figure;
stackedplot(tt);
title('Data Trends Over Time');
else
% Create standard plot
figure;
plot(cleanData.Variables);
legend(cleanData.Properties.VariableNames);
title('Data Trends');
end
Create an app that allows users to upload an image and apply filters
Agent creates App Designer code or programmatic UI:
function imageFilterApp
% Create figure and components
fig = uifigure('Name', 'Image Filter App');
g = uigridlayout(fig, [3 2]);
% Load button
btnLoad = uibutton(g, 'Text', 'Load Image', ...
'ButtonPushedFcn', @loadImage);
btnLoad.Layout.Row = 1;
btnLoad.Layout.Column = 1;
% Filter dropdown
dd = uidropdown(g, 'Items', {'None', 'Gaussian', 'Median', 'Edge'});
dd.Layout.Row = 1;
dd.Layout.Column = 2;
dd.ValueChangedFcn = @applyFilter;
% Image axes
ax = uiaxes(g);
ax.Layout.Row = [2 3];
ax.Layout.Column = [1 2];
img = [];
function loadImage(~, ~)
[file, path] = uigetfile({'*.jpg;*.png;*.bmp', 'Image Files'});
if file
img = imread(fullfile(path, file));
imshow(img, 'Parent', ax);
end
end
function applyFilter(~, ~)
if isempty(img)
return;
end
filtered = img;
switch dd.Value
case 'Gaussian'
filtered = imgaussfilt(img, 2);
case 'Median'
filtered = medfilt3(img);
case 'Edge'
if size(img, 3) == 3
gray = rgb2gray(img);
else
gray = img;
end
filtered = edge(gray, 'Canny');
end
imshow(filtered, 'Parent', ax);
end
end
By default, the toolkit configures your agent globally. To use project-specific configuration with the MATLAB-based installer:
setupAgenticToolkit('Scope', 'project')
To connect the MCP server to an already-running MATLAB session:
setupAgenticToolkit('SessionMode', 'existing')
Or manually configure the MCP server with --matlab-session-mode=existing flag.
Ask your agent:
What version of MATLAB is running? List the installed toolboxes.
Expected output shows MATLAB version and all installed toolboxes.
matlab -batch "disp(version)"
If skills aren't loading:
claude plugin list
When evaluate_matlab_code fails:
check_matlab_code first for syntax validationOn Windows, if installation fails with permission errors:
On macOS/Linux:
ls -la ~/.config/ (or agent-specific config location)which matlabIf detect_matlab_toolboxes returns incomplete results:
% Run in MATLAB directly to verify
ver
Compare output with agent results. Missing toolboxes may need licensing or installation.
Always review tool calls before execution. The MCP server executes code with full MATLAB privileges:
evaluate_matlab_code calls before approvalrun_matlab_file, save, load)Licensing: MCP servers must comply with MathWorks Software License Agreement and cannot be shared by multiple users. Contact MathWorks for multi-user deployments.
The toolkit respects these environment variables:
MATLAB_ROOT: Override MATLAB installation path detectionMCP_SERVER_LOG_LEVEL: Set logging verbosity (debug, info, warn, error)Reference in configuration:
export MATLAB_ROOT=/Applications/MATLAB_R2024b.app