VMVM

How do you want to use VM?

Choose Agent to start the local VM and model tools, or Human to connect to an agent that is already running its AI model.

About this project — architecture, components & licenses

Preparing VMVM…

Connect to an agent instead

VMVM

loading v86…

Local AI model: not configured

AutoBro: not connected

Voice: ready to load

VM runtime fault detected The VM was kept open so the error can be inspected.

Preparing voice input Loading Moonshine Tiny…

Loading AI model 0%

Preparing model file…

Settings

VMVM Image

Select a cumulative guest image. Changing it restarts VMVM; each image has an independent guest filesystem.

Image

AI Model

Load a LiteRT-LM model into this browser or clear all cached models.

Loaded model

Download Gemma 4 E2B (2.0 GB) Download Gemma 4 E4B (3.0 GB) Download Gemma 4 12B (6.0 GB)

More info

Learn how LiteRT-LM runs fast generative AI directly on the device: Blazing-fast on-device GenAI with LiteRT-LM.

Cloud AI providers

Local WebGPU remains the direct default. Add a cloud provider for optional per-agent or command-line selection. Keys stay in this tab unless you explicitly retain them.

Saved provider

Name API type Model Base URL API key Retain key in this browser

Rig default Zerostack default vmlang default vmmastra default

AutoBro

Connect browser automation or forget the current pairing.

Download AutoBro Chrome extension 2026.07.29.8 Download Guidewire PolicyCenter skills 1.0.0 Download Windows networking helper

AutoBro is not installed or enabled.

Download AutoBro 2026.07.29.8 and extract the ZIP.
Open chrome://extensions in Chrome or Chromium.
Enable Developer mode, then select Load unpacked.
Select the extracted folder containing manifest.json.
Return here and select Connect again.

Voice

Load Moonshine Tiny locally and insert transcription at the current prompt.

Assets transfer

Import or export a complete project archive, or transfer one regular guest file up to 8 MiB.

Remote agent

Run this model as an agent or chat with a paired agent over WebRTC.

Diagnostics

Local capability and runtime metrics only. Project files, AI prompts, and pairing credentials are excluded. A retained VM fault may include bounded serial context and the last shell command.

Help
Open the VM command reference, including local and cloud LLM connections.

Remote agent

Starting secure WebRTC agent…

Pairing key

Connection and model activity

Paste into VM shell

Your browser did not allow direct clipboard access. Paste the text below, then select Insert.