Antigravity + MATLAB is wicked good

著者 Yann Debray, December 24, 2025

126 ビュー (過去 30 日間) | 0 いいね | 0 コメント

Google released a new agentic development environment called Antigravity. And it works great with MATLAB. (It's wicked as they say in Boston!)

If you are looking to defy gravity, here is a fun use case for an app leveraging the power of MATLAB and Antigravity:

This article will look at three different components, that were built using Antigravity + MATLAB:

Component	Description
uihtml + React workflow	Modern web frameworks embedded seamlessly in MATLAB desktop apps
Flight data replay	Post-process and visualize trajectories with gravityFlightDashboard
Defying Gravity Toolbox	A fully-packaged MATLAB app with 3D graphics powered by Three.js

The AI Copy & Paste Problem (Revisited for apps)

Regular readers will remember the core problem that led us to release the MATLAB MCP Core Server: the tedious copy-paste loop between your favorite AI chat and MATLAB. Write code in ChatGPT or Claude, copy it, paste it in MATLAB, get an error, copy the error, paste it back... iterate.

The MCP server solved that for MATLAB code. But what about apps?

Building a MATLAB app with modern 3D graphics used to require deep knowledge of app building in MATLAB, plus OpenGL graphics. Now, with uihtml, you can embed any web technology—React, Three.js, D3.js—directly inside a MATLAB figure. The catch? You need to know web development too.

That is where AI agents come in. I described what I wanted in natural language, and Antigravity (with the MATLAB MCP Core Server connected) helped me scaffold the whole thing.

Connect MATLAB to Antigravity

Let's start by connecting MATLAB to Antigravity. There are two steps you can take, quite similar to how you would in VSCode (Antigravity is a fork of VSCode):

Install the MATLAB extension
Set up the MATLAB MCP Core Server

Navigate to the extension side-panel, and search for MATLAB:

You will find the MCP Server setup screen in the three dots on the top right of the Agent side panel (slightly different from VSCode).

Once your MATLAB is connected to Antigravity, you can ask the following prompt:

Generate a MATLAB App with 2 windows: One web frontend implementing a flight simulator in React and ThreeJS that enables to record the trajectory in a CSV file. This web frontend will be integrated in a MATLAB uihtml component. The second window is a dashboard post-processing the trajectories with MATLAB plots. There is a button that enables to open the simulator from this view.

Testing web frontend before integration in uihtml

MATLAB Apps can integrate HTML UI components through the use of the function uihtml. One great combo is to use the embedded chrome browser in Antigravity to test the interactions of the web frontend before integrating it into your app:

Let's Walk Through the Generated Code

The Main App Entry Point

The core of gravityFlightSimulator.m is surprisingly simple:

function gravityFlightSimulator()

% Create main figure with appropriate size

 fig = uifigure('Name', 'Defying Gravity', ...

'Position', [100 100 1200 800], ...

'Color', [0 0 0], ...

'AutoResizeChildren', 'off');

% Get the path to the HTML file and icon

 appDir = fileparts(mfilename('fullpath'));

 htmlFile = fullfile(appDir, 'dist', 'index.html');

 iconFile = fullfile(appDir, 'dist', 'broom.png');

% Set the app icon (yes, it's a broom)

 fig.Icon = iconFile;

% Create uihtml component filling the entire figure

 h = uihtml(fig, 'Position', [0 0 fig.Position(3) fig.Position(4)]);

 h.HTMLSource = htmlFile;

% Handle figure resize to keep uihtml fullscreen

 fig.SizeChangedFcn = @(src, ~) resizeHTML(src, h);

end

That's it. About 20 lines of MATLAB to embed a full 3D flight simulator.

The Flight Controls

The web app (built with React and Three.js) provides the flight physics:

Control	Action
W/S	Throttle Up/Down
Up/Down	Pitch Up/Down
Left/Right	Roll Left/Right

And you can switch between a plane and a witch on a broomstick (hence the icon). Because why not?

Post-Processing with gravityFlightDashboard

The app can export flight data to CSV. Then comes the MATLAB part—analyzing trajectories with gravityFlightDashboard:

gravityFlightDashboard('gravity-flight-2025-12-07.csv')

This opens a replay dashboard with:

3D trajectory visualization with heading arrows
Altitude vs. time plot
Play/Pause/Reset controls with a timeline slider
Real-time flight metrics: speed, altitude, pitch, roll

% The dashboard using MATLAB graphics

plot3(pathAxes, position(:,1), position(:,2), position(:,3), ...

'Color', [0.7 0.7 0.7], 'LineWidth', 1.25);

craftMarker = plot3(pathAxes, position(1,1), position(1,2), position(1,3), ...

'o', 'MarkerFaceColor', [0.95 0.3 0.2], 'MarkerEdgeColor', 'k', ...

'MarkerSize', 8);

This is the beauty of the hybrid approach: use web tech for the interactive 3D simulation, use MATLAB for post-processing and analysis.

If you are interested in more advanced flight data analysis, I recommend you take a look at our UAV toolbox.

Packaging as a Toolbox

One of my favorite parts: you can package this as a proper MATLAB toolbox (.mltbx file) that anyone can install with one click.

The "Gravity" toolbox appears right alongside your other add-ons.

Install it like this:

matlab.addons.install('Gravity.mltbx')

Then just run:

gravityApp

The toolbox bundles everything: the MATLAB code, the web assets, the broom icon. Users don't need to know anything about React or Three.js—they just install and fly.

Calling the Wizard of Apps

For the next post of this series on Agentic App Building, I will turn to colleague Pax, who I call the "Wizard of Apps":

Next episode: Rick Paxson aka "Pax"

Rick Paxson is the Director of Computational Biology at MathWorks. He is an experienced programmer and an early adopter of AI assisted coding. I always turned to him for advices on how to best leverage AI in software development.

He developed some particular "skills" that he shared with Ned and I to build MATLAB Apps with a web front end. If you are too impatient, you can get a sneak peek into the recipe we will use: skills/skills/matlab-uihtml-app-builder/SKILL.md

In the meantime, I wish you a wicked Christmas🎅

PS: Check out the code and have fun defying gravity🧹

MATLAB-AI-Blog/antigravity-plus-matlab-is-wicked-good at main · mathworks/MATLAB-AI-Blog