Skip to content

Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.

3.6 KB
Raw Blame

three.ws Walk Avatar (Chrome Extension)

Walk your three.ws avatar on any website you visit, with optional page narration.

Install (development / load-unpacked)

  1. Build:
    npm run build:extension
    
  2. Open chrome://extensions
  3. Enable Developer mode (toggle top-right)
  4. Click Load unpacked
  5. Select dist/extension/

Usage

  • Click the extension icon in the toolbar to open the popup
  • Sign in with your three.ws account to access your avatars
  • Select an avatar and toggle "Enable on this site"
  • The avatar appears floating in the corner of the page and re-mounts automatically after full page loads on tabs where you enabled it
  • Hover the avatar and drag the handle at the top to reposition it
  • Click the speaker button to mute narration, or the close button to hide the avatar on that page without disabling the extension
  • Open the options page (right-click the icon, then Options) for position, size, speed, theme, narration voice, and per-site allow/blocklists

Build for Web Store

npm run build:extension:prod

Outputs a minified build plus dist/extension-<version>.zip (version stamped from the root package.json) ready to upload to the Chrome Web Store. The full submission checklist lives in RELEASE.md.

Structure

extensions/walk-avatar/
├── manifest.json         Manifest V3 config
├── background.js         Service worker: settings, auth capture, injection, tab re-mount
├── content.js            Content script: avatar iframe mount, drag, toolbar
├── content-narrator.js   Page-section narrator (TTS via three.ws)
├── vendor/readability.js Vendored Readability for article extraction
├── popup.html/.js/.css   Extension popup: avatar picker + per-tab toggle
├── options.html/.js      Full settings page
├── icons/                Generated by scripts/build-extension-icons.mjs
├── store-assets/         Chrome Web Store listing images
├── PERMISSIONS.md        Per-permission justifications for store review
├── PRIVACY.md            Privacy policy source (live at three.ws/extension/privacy)
├── TERMS.md              Terms source (live at three.ws/extension/terms)
└── RELEASE.md            Web Store release checklist

The build is scripts/build-extension.mjs (esbuild): background.js bundles as an ESM service worker, everything else as self-contained IIFEs, and static files (manifest, HTML, CSS, icons, vendor) are copied alongside. Output lands in dist/extension/.

Permissions

  • storage: stores avatar selection, settings, and the session token
  • activeTab: reads the current tab URL for the site label in the popup
  • tabs: re-mounts the avatar after full page loads on enabled tabs, relays live setting changes, and opens sign-in/create pages
  • scripting: injects the content scripts when the user enables the avatar
  • host_permissions: <all_urls>: required to inject into any site the user chooses

Full store-review justifications: PERMISSIONS.md.

Privacy

User data stays on-device in chrome.storage. The only outbound requests, all to three.ws:

  • GET /api/avatars/mine, GET /api/avatars/featured, GET /api/avatars and thumbnails: avatar lists for the pickers (sign-in required for private avatars)
  • GET /api/tts/voices: the narration voice catalog shown in options
  • POST /api/tts/speak: synthesizes narration audio, only when narration is explicitly enabled
  • The avatar iframe itself loads three.ws/walk-embed to render the 3D scene

Policy source files: PRIVACY.md, TERMS.md.