Getting Started with RunMat

Get up and running in minutes. Try RunMat in your browser with no installation, or install the CLI for the terminal and local scripts. Have existing .m files? Paste them in the sandbox and run them. The built-in agent can suggest fixes for unsupported functions.

Launch Browser App

No installation required. Works in Chrome, Edge, Firefox, and Safari.

Choose your path

RunMat in the browser gives you a full IDE—editor, file explorer, console, and live plots—all running locally. No installation. Visit runmat.com/sandbox to start. Works in Chrome, Edge, Firefox, and Safari. For GPU acceleration, use a browser with WebGPU (Chrome 113+, Edge 113+, Safari 18+, Firefox 139+).

The interface

Four main areas: Sidebar (file tree, + to add files), Editor (code, Cmd/Ctrl+S to save), Runtime Panel (Run, Figures, Console, Variables), and Agent Panel (chat, reviewable diffs). Panels are resizable.

1Run the demo

Try this example — paste it into a new file or run it directly:

a = 0:pi/100:2*pi;
b = sin(a);

plot(a, b);

Click the purple ▶ Run demo.m button in the Runtime Panel. A Figure tab appears with the sine wave; the Variables tab shows workspace variables with types, shapes, and CPU/GPU residency.

2Edit and experiment

Try changing the plot — for example, swap sin for cos and re-run:

b = cos(a);

Use Ctrl+Enter (Windows/Linux) or Cmd+Enter (macOS) to run without clicking the button.

3Create your own script

Click + in the sidebar → Create file (.m) → name it (e.g. my_script.m). The editor supports standard MATLAB syntax; large arrays get automatic GPU acceleration via WebGPU.

4Bring your existing scripts

Paste an existing .m file and run it. If something fails — an unsupported function or a toolbox-specific call — open the built-in agent and ask it to rework the code. It will propose reviewable edits you can accept or discard. See Agent-assisted migration for details.

Sandbox storage: Files live in your browser tab. No account required — code execution is local. Using the built-in agent sends context to the configured LLM provider. Files are cleared when you close or refresh the tab. Sign in free for 100 MB cloud storage with automatic versioning, or use the CLI for local persistence.

Compare surfaces

Feature	CLI	Sandbox No account	Sandbox + App Signed in
Works without install
Account required			Yes (free)
GPU acceleration	Native (Metal, Vulkan, DX12)	WebGPU (browser-throttled)	WebGPU (browser-throttled)
Interactive IDE
Interactive plotting	GUI window	In-editor	In-editor
Variable inspector
File storage	Local filesystem	In-memory (cleared on tab close)	App (persists across sessions)
File versioning	No (use git)		Automatic
Project sharing			Paid plans
Jupyter support
Offline support
Built-in agent