Puppeteer MCP Server
Enables AI agents to automate browser interactions including navigation, content extraction, form filling, screenshots, and JavaScript execution across multiple tabs using Puppeteer.
README
Puppeteer MCP Server
A Model Context Protocol (MCP) server that provides browser automation capabilities through Puppeteer. This server enables AI agents to interact with web pages, take screenshots, execute JavaScript, and perform various browser operations.
Features
- Multi-tab Support: Manage multiple browser tabs with unique IDs
- Comprehensive Tools: 27 tools for navigation, interaction, content extraction, and more
- Dual Transport: Supports both stdio (for Claude Desktop/Code) and HTTP transports
- Result Types: Consistent error handling with structured Result types
Installation
npm install
npm run build
Usage
Stdio Mode (Default)
For use with Claude Desktop or Claude Code:
npm start
# or
node dist/index.js
HTTP Mode
For remote or containerized deployments:
npm start -- --http
# or
node dist/index.js --http --port=3000
Claude Desktop Configuration
Add to your Claude Desktop config:
{
"mcpServers": {
"puppeteer": {
"command": "node",
"args": ["/path/to/puppeteer-mcp/dist/index.js"]
}
}
}
Available Tools
Tab Management
| Tool | Description |
|---|---|
list_tabs |
List all open browser tabs |
new_tab |
Create a new tab (optionally with URL) |
close_tab |
Close a tab |
switch_tab |
Switch to a different tab |
Navigation
| Tool | Description |
|---|---|
navigate |
Navigate to a URL |
reload |
Reload the current page |
go_back |
Navigate back in history |
go_forward |
Navigate forward in history |
Interaction
| Tool | Description |
|---|---|
click |
Click an element |
fill |
Fill a text input |
select |
Select dropdown option(s) |
hover |
Hover over an element |
focus |
Focus an element |
Input
| Tool | Description |
|---|---|
keyboard |
Press keyboard keys |
mouse |
Perform mouse actions |
scroll |
Scroll the page or element |
Content
| Tool | Description |
|---|---|
evaluate |
Execute JavaScript |
get_content |
Get page/element HTML or text |
query_selector |
Get element information |
Waiting
| Tool | Description |
|---|---|
wait_for_selector |
Wait for element to appear |
wait_for_navigation |
Wait for navigation |
wait |
Wait for specified time |
Media
| Tool | Description |
|---|---|
screenshot |
Capture screenshot |
pdf |
Generate PDF |
Cookies
| Tool | Description |
|---|---|
get_cookies |
Get cookies |
set_cookies |
Set cookies |
delete_cookies |
Delete cookies |
Tool Parameters
All tools that operate on pages accept an optional tabId parameter. If not specified, the active tab is used.
Example: Navigate and Take Screenshot
// Navigate
{ "name": "navigate", "arguments": { "url": "https://example.com" } }
// Take screenshot
{ "name": "screenshot", "arguments": { "fullPage": true, "format": "png" } }
Example: Multi-tab Workflow
// Create new tab
{ "name": "new_tab", "arguments": { "url": "https://site-a.com" } }
// Returns: { "id": "tab_abc123", "url": "https://site-a.com", ... }
// Create another tab
{ "name": "new_tab", "arguments": { "url": "https://site-b.com" } }
// Returns: { "id": "tab_def456", "url": "https://site-b.com", ... }
// List all tabs
{ "name": "list_tabs", "arguments": {} }
// Interact with specific tab
{ "name": "click", "arguments": { "selector": "button", "tabId": "tab_abc123" } }
Environment Variables
| Variable | Description | Default |
|---|---|---|
PORT |
HTTP server port | 3000 |
HEADLESS |
Run browser headless | true |
TIMEOUT |
Default operation timeout (ms) | 30000 |
Development
# Install dependencies
npm install
# Run in development mode
npm run dev
# Type check
npm run typecheck
# Run tests
npm test
# Build
npm run build
Architecture
src/
├── index.ts # Entry point, transport setup
├── server.ts # MCP server configuration
├── browser.ts # Browser lifecycle management
├── tabs.ts # Multi-tab state management
├── types.ts # TypeScript interfaces
├── errors.ts # Result types and error handling
├── schemas.ts # Zod validation schemas
└── tools/
├── tab-tools.ts # Tab management tools
├── navigation.ts # Navigation tools
├── interaction.ts # Click, fill, select, etc.
├── content.ts # Evaluate, get content
├── waiting.ts # Wait tools
├── media.ts # Screenshot, PDF
├── cookies.ts # Cookie management
└── input.ts # Keyboard, mouse, scroll
License
MIT
puppeteer-mcp
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。