Browser Agent Extension

中文文档 | English

AI-powered browser agent extension — control tabs, windows, bookmarks, history, downloads, cookies, and more with natural language.

Features

• 🧠 Natural Language — Control your browser via a sidebar chat interface
• 🗂️ Full Browser Management — Tabs, windows, groups, bookmarks, history, downloads, cookies, sessions, and more
• 🔒 Safety Guardrails — Sensitive operations require user confirmation to prevent accidents
• 🎯 Cross-Browser — Supports both Chrome (MV3) and Firefox
• 🧩 Extensible Tool System — Phased registration with Expert Mode support
• 🎤 Voice Input — Voice interaction support (in development)

Quick Start

# Install dependencies
npm install

# Dev mode (Chrome)
npm run dev:chrome

# Dev mode (Firefox)
npm run dev:firefox

# Build
npm run build

# Build all browser versions
npm run build:all

# Run tests
npm run test:run

# Lint
npm run lint

# Type check
npm run typecheck

Tech Stack

Category	Technology
Framework	WXT — Next-gen Web Extension Framework
UI	React 18 + TypeScript
Styling	TailwindCSS v4
Build	Vite
Testing	Vitest + Playwright + Testing Library
Lint	ESLint + Prettier

Project Structure

src/
├── adapters/        # Chrome/Firefox browser adapter layer
├── agent/           # Agent loop, context, system prompts
├── background/      # Service Worker logic
├── chat/            # Chat logic
├── content/         # Content Script
├── conversation/    # Conversation management
├── entrypoints/     # WXT entrypoints
│   ├── background.ts
│   ├── content.ts
│   └── sidepanel/   # Sidebar UI (App.tsx + components)
├── guardrail/       # Safety guardrails, confirmation
├── provider/        # LLM & voice recognition clients
├── registry/        # Tool registration
├── tools/           # Browser API tools
│   ├── tabs/        # Tab operations
│   ├── bookmarks/   # Bookmark operations
│   ├── cookies/     # Cookie operations
│   ├── downloads/   # Download management
│   ├── history/     # History
│   ├── sessions/    # Session snapshots
│   ├── windows/     # Window management
│   ├── tabgroups/   # Tab groups
│   ├── page/        # Page content reading
│   └── misc/        # Miscellaneous
├── types/           # Type definitions
└── shared/          # Shared utilities

Build Outputs

Browser	Manifest	Output
Chrome	MV3	.output/chrome-mv3/
Firefox	MV2	.output/firefox-mv2/

# Package as zip
npm run zip

# Full build + verification
npm run pre-release

Capabilities

Default: tabs, windows, tabGroups, bookmarks, history, downloads, sessions, page content, cookies, storage, clipboard, notifications, contextMenus, sidePanel (Chrome), alarms

Expert Mode (manual enable): proxy, privacy, management, debugger, webRequest, declarativeNetRequest, nativeMessaging, identity

Development Guide

# Run tests (watch mode)
npm run test

# Run specific tests
npx vitest run src/tools/tabs

# E2E tests
npx playwright test

# Format code
npm run format

Documentation

• PRD
• Design Docs
• Review Records
• Issue Diagnosis