Browser Bridge MCP
MCP server that gives AI coding assistants direct access to the browser — navigate, click, fill forms, run JavaScript, take screenshots, and read page content.
README
Browser Bridge MCP
MCP server that gives AI coding assistants direct access to the browser — navigate, click, fill forms, run JavaScript, take screenshots, and read page content.
MCP Client ←(stdio)→ mcp-browser-bridge ←(WebSocket :7483)→ Browser Extension ←(Chrome APIs)→ Browser
Works with any MCP-compatible client: Claude Code, Cursor, Windsurf, Cline, and others.
Quick Start
1. Install the extension
Install from the Chrome Web Store, or load manually:
- Download or clone this repo.
- Open
chrome://extensions, enable Developer mode. - Click Load unpacked and select the
extension/folder.
Works in any Chromium browser: Chrome, Brave, Edge, Arc, Vivaldi, Opera.
2. Add to your MCP client
<details> <summary><strong>Claude Code</strong></summary>
Run:
claude mcp add browser-bridge -- npx -y mcp-browser-bridge
Or add to .mcp.json:
{
"mcpServers": {
"browser-bridge": {
"command": "npx",
"args": ["-y", "mcp-browser-bridge"]
}
}
}
</details>
<details> <summary><strong>Cursor</strong></summary>
Add to .cursor/mcp.json:
{
"mcpServers": {
"browser-bridge": {
"command": "npx",
"args": ["-y", "mcp-browser-bridge"]
}
}
}
</details>
<details> <summary><strong>Windsurf</strong></summary>
Add to ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"browser-bridge": {
"command": "npx",
"args": ["-y", "mcp-browser-bridge"]
}
}
}
</details>
<details> <summary><strong>Other MCP clients</strong></summary>
Use npx -y mcp-browser-bridge as the server command in your client's MCP configuration. The server communicates over stdio.
</details>
3. Verify
Start your MCP client. The extension popup should show a green Connected indicator. Call the browser_status tool to confirm.
Available Tools
All tools accept an optional tabId parameter. When omitted, they target the active tab.
| Tool | Description |
|---|---|
browser_status |
Check if the extension is connected and get active tab info |
browser_navigate |
Navigate a tab to a URL |
browser_screenshot |
Capture a screenshot of the visible area of a tab |
browser_evaluate |
Execute JavaScript in a tab and return the result |
browser_click |
Click an element by CSS selector |
browser_fill |
Fill a form field by CSS selector (React-compatible) |
browser_get_content |
Get the text or HTML of a page or element |
browser_get_tabs |
List all open browser tabs |
browser_get_console |
Get captured console log entries from a tab |
browser_wait_for |
Wait for a CSS selector to appear on the page |
browser_send_message |
Send a custom message to the extension |
Included Prompts
The server ships two MCP prompts that provide guided workflows to any connected client:
| Prompt | Description |
|---|---|
browse |
General-purpose browser interaction — navigating, clicking, filling forms, screenshotting, running JS. Accepts an optional task argument. |
qa-runner |
Structured QA checklist execution — drives the browser through test scenarios defined in markdown and reports pass/fail results. Accepts an optional checklist path. |
In Claude Code, these appear as slash commands: /browser-bridge:browse and /browser-bridge:qa-runner.
Configuration
The WebSocket port defaults to 7483. To change it:
- Server side: Set the
BRIDGE_WS_PORTenvironment variable in your MCP config. - Extension side: Change the port in the extension popup and click Reconnect.
Security
- WebSocket binds to
127.0.0.1only — no network exposure. - Connections restricted to browser extension origins.
- Single extension client at a time.
Troubleshooting
Extension shows "Disconnected" Check that the MCP server is running and the port is free:
# macOS/Linux
lsof -i :7483
# Windows
netstat -aon | findstr :7483
Tools return "No extension connected" Open the extension popup and click Reconnect.
Screenshots fail The target tab must be visible and focused. Background or minimized tabs cannot be captured.
browser_evaluate returns unexpected results
Results must be JSON-serializable. Promises return {} — use synchronous expressions or await inside an IIFE and return a plain value.
Development
git clone https://github.com/adbarc92/mcp-browser-bridge.git
cd mcp-browser-bridge
npm install
npm run build
npm test # 53 tests via Vitest
Load the extension locally via chrome://extensions → Load unpacked → select extension/.
The repo includes a .mcp.json configured for local development.
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。