Repository navigation
v0.1.0 — First release
J.A.R.V.I.S. v0.1.0 — first release
A browser voice assistant with an Iron Man holographic interface. Say "Hey Jarvis" and he wakes, listens, and does real things through your tools. The face is a React + Vite + Three.js + custom GLSL page. The brain is Claude Code, run headless through the Claude Agent SDK.
Highlights
- Wake word and hands-free voice loop: local voice-activity detection, barge-in (speak while he talks and he stops), and automatic recovery if speech recognition stalls.
- Holographic reactor UI with a heads-up display, rendered with Three.js and GLSL.
- Claude Code as the brain: reuses your existing Claude Code login, so there is no API key and no separate cloud bill.
- Real tools through MCP: web search, image generation, phone control, mail and more, run by a local bridge that spawns your MCP servers.
- Permission gate: effectful tools are off by default and enabled only with
JARVIS_ALLOW_WRITES=1. - Optional ElevenLabs add-on for a better voice and sharper transcription. Without it, JARVIS uses the browser's own speech.
Requirements
- Claude Code, installed and logged in
- Node.js 20+
- Chrome or Edge in a real browser window (embedded preview panes block the microphone)
Quick start
npm install
npm run setup # checks your environment
npm start # runs the brain and the face togetherOpen http://localhost:5173 in Chrome, click INITIALISE, and say "Hey Jarvis".
See the README for architecture and configuration.