M-GPUX — Modal GPU Orchestrator

Provision GPU workloads, manage multi-profile Modal accounts, and track cloud billing — all without leaving your editor.

Features

• GPU Hub Wizard — Guided multi-step flow to launch Jupyter, Python scripts, web shells, or vLLM inference servers on any NVIDIA GPU
• Multi-Account Management — Add, switch, and remove Modal profiles from the sidebar
• Live Billing — See remaining credits per account directly in the sidebar; hover for detailed breakdown
• Hardware Probing — Inspect GPU/CPU/RAM metrics on any Modal container
• Native VS Code UI — Output Channels for logs, progress notifications with cancel, clickable URL pop-ups when tunnels are ready

Requirements

Installation

# From VSIX file
code --install-extension m-gpux-2.5.0.vsix

Or in VS Code: Ctrl+Shift+P → Extensions: Install from VSIX...

After installing, run Reload Window (Ctrl+Shift+P → Reload Window).

Getting Started

1. Open the Sidebar

Click the GPU chip icon on the Activity Bar (left edge). Two panels appear:

2. Add Your First Account

Click the + button on the Accounts panel header, or run Ctrl+Shift+P → M-GPUX: Add Account.

You can add credentials in two ways:

• Paste shortcut — Paste a full modal token set --token-id ak-... --token-secret as-... --profile=myname command. The extension parses it automatically.
• Manual entry — Leave the input empty and fill in Token ID, Token Secret, and Profile Name step by step.

The first profile is automatically set as active.

3. Launch a GPU Workload

Click GPU Hub in Quick Actions, or run Ctrl+Shift+P → M-GPUX: Open GPU Hub.

The wizard walks you through four steps:

Step 1 — Select Profile

Choose a Modal account. AUTO uses the currently active profile.

Step 2 — Choose GPU

Step 3 — Choose Application

Step 4 — Configure & Launch

• For Jupyter and Python: optionally use requirements.txt and set exclude patterns for file upload.
• The extension generates a modal_runner.py script and opens it in the editor for review.
• Click Launch → the process runs in the background with a progress notification.
• When a tunnel URL is detected, a notification pops up with Open in Browser and Copy URL buttons.
• The temp script is cleaned up automatically when the process exits.

Account Management

Status Bar

The bottom-left of VS Code shows ☁ M-GPUX: <profile-name>. Click it to switch profiles.

Billing

• Sidebar — Each account shows remaining credit (e.g. $27.50 left). Hover for a used/remaining breakdown.
• Billing Usage — Click in Quick Actions to run m-gpux billing usage --all and see a detailed cost table.
• Billing Dashboard — Opens modal.com/settings/usage in your browser.

All Commands

Open the Command Palette (Ctrl+Shift+P) and type M-GPUX:

Configuration

The extension reads and writes ~/.modal.toml directly — the same file used by the Modal CLI. No separate config needed.

Example ~/.modal.toml:

[personal]
token_id = "ak-xxxx"
token_secret = "as-xxxx"
active = true

[work]
token_id = "ak-yyyy"
token_secret = "as-yyyy"

Quick Workflow

1. Open VS Code with your Python project
2. Click the M-GPUX icon on the sidebar
3. Add an account if you haven't already (+ button)
4. Click "GPU Hub" → pick a GPU → pick Jupyter Lab
5. Review the generated script → click Launch
6. Notification pops up with your Jupyter URL → Open in Browser
7. Code on a remote GPU with your workspace files ready at /workspace

Troubleshooting