BrowserOS Server

The automation engine inside BrowserOS. This MCP server powers the built-in AI agent and lets external tools like claude-code or gemini-cli control your browser.

How It Works

When BrowserOS launches, this server starts automatically in the background. The built-in agent UI connects to it, giving you a sidepanel where you can chat with AI and automate browser tasks.

BrowserOS launches
       ↓
MCP Server starts (this repo)
       ↓
Agent extension connects
       ↓
You can chat and automate tasks

You can also connect external MCP clients to control BrowserOS from your terminal or other AI tools.

Architecture

The server exposes browser automation tools through two channels:

• Controller tools - High-level commands via a Chrome extension (navigation, clicks, screenshots, tabs, history, bookmarks)
• CDP tools - Low-level access via Chrome DevTools Protocol (network inspection, console logs, emulation)

apps/
├── server/          # MCP server (HTTP + WebSocket)
└── controller-ext/  # Chrome extension for browser control

For Developers

# Install dependencies
bun install

# Start the server (for local development)
bun run start

# Run tests
bun run test              # Unit tests
bun run test:cdp          # CDP tests (needs browser running)
bun run test:all          # Everything

# Lint and type check
bun run lint
bun run typecheck

# Build for production
bun run dist:server       # All platforms
bun run dist:ext          # Extension

Using with External MCP Clients

Once BrowserOS is running, you can connect any MCP client to http://127.0.0.1:<port>/mcp. See our docs for setup with claude-code and gemini-cli.

Related Repos

• BrowserOS - The Chromium fork (main product)
• BrowserOS Agent - The built-in sidepanel extension for chatting with AI

License

AGPL-3.0-or-later