Your AI copilot for the web — a Chrome/Edge browser extension that brings GitHub Copilot directly into your browser.
- Side Panel Chat: Persistent chat interface accessible via toolbar icon or Ctrl+Shift+Y / ⌘+Shift+Y
- Page Understanding: Extract content, structure, forms, tables, and links from any web page
- Browser Automation: Click, type, fill forms, navigate, and interact with web pages
- Screenshots: Capture visible viewport or specific elements
- Visual Feedback: Highlight elements on the page during agent interactions
- Session History: Persistent chat sessions stored locally
- Dark/Light Mode: Follows your system theme preference
- Node.js 18+
- Chrome or Edge browser
- GitHub Copilot CLI (
@github/copilot)
-
Install dependencies:
npm install
-
Build the extension:
npm run build
-
Load the extension:
- Go to
chrome://extensions(oredge://extensions) - Enable "Developer mode"
- Click "Load unpacked"
- Select the
dist/directory
- Go to
-
Register the native messaging host:
# macOS/Linux scripts/register-host.sh <your-extension-id> # Windows scripts\register-host.bat <your-extension-id>
Your extension ID is shown on the extensions page after loading.
-
Restart your browser and click the Copilot icon in the toolbar.
npm run dev # Start Vite dev server with HMR
npm run build # Production buildSide Panel (React) ←→ Background Service Worker ←→ Native Messaging Host (Node.js) ←→ Copilot CLI
↕
Content Scripts (DOM access)
The agent has access to 27+ tools across categories:
- Page Content: get_page_content, get_page_html, get_page_structure, query_selector, get_page_tables, get_page_links, get_page_forms
- Visual: capture_screenshot, highlight_element
- Interaction: click_element, type_text, fill_form, select_option, scroll_page, press_key
- Navigation: navigate_to, go_back, go_forward, open_tab, close_tab, switch_tab, reload_page
- Utility: wait_for_element, execute_javascript
MIT