Selenium Mcp

io.github.pshivapr由 GitHub 账号 pshivapr 发布v0.4.6更新于 Oct 24, 2026

Selenium Tools for MCP

已验证STDIO仅桌面网页搜索与抓取浏览器自动化

概览

AI 生成的概览

一个本地 MCP 服务器，把 Selenium WebDriver 暴露给助手，用于驱动真实浏览器：导航页面、点击输入、管理 Cookie 与窗口并截图。

可以做什么
打开、导航、刷新、调整大小和关闭浏览器会话，并读取当前 URL、标题和页面源码，支持 Chrome、Firefox、Safari 和 Edge。查找元素并与之交互：点击、输入、清空、悬停、双击、右键、拖放、上传文件，以及读取文本或属性。管理 Cookie 和窗口，包括在窗口与 iframe 之间切换。执行高级操作，如运行 JavaScript、截图、选择下拉项、按键、滚动和提交表单。检查元素状态（是否显示、启用、选中），并等待元素或页面状态。
适用场景
适合需要助手执行真实浏览器自动化的场景：端到端网页测试、填写表单、抓取需要真实浏览器的页面，或根据观察到的流程生成 Selenium 测试代码。
运行要求
通过 stdio 在本地运行，通常用 npx 从 npm 包启动。需要 Node.js 18 或更高版本，以及已安装的浏览器（Chrome、Firefox、Safari 或 Edge）；WebDriver 会自动管理。清单声明了一个可选的 API 密钥变量，但 README 未说明其用途。
安装前请注意
助手可控制真实浏览器：能提交表单、点击按钮、上传本地文件并在页面中执行任意 JavaScript，可能修改数据或在已登录的网站上触发操作。Cookie 工具可读取和修改会话 Cookie，避免用于敏感账户。截图和页面源码可能包含私密内容。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Selenium Mcp，将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Desktop only，通过 STDIO。 STDIO 服务会启动本地进程，因此需要 SourceWeft 桌面宿主。

其他 MCP 客户端

参照 仓库 中的启动说明。

README

[Add to Cursor] [Add to VS Code] [Add to Claude] [Add to ChatGPT] [Add to Codex] [Add to Gemini]

Selenium MCP Server

[npm version]

[npm downloads]

[GitHub issues]

[smithery badge]

This is a server implementation that bridges the gap between MCP clients (AI assistants) and Selenium WebDriver. It exposes Selenium WebDriver's functionalities as MCP tools, allowing AI models to utilize them for tasks like:

  • Browser management (launching, navigating, closing browsers)
  • Element interaction (clicking, typing, finding elements)
  • Web scraping and automated testing
  • Advanced operations like screenshots, cookie management, and JavaScript execution

In essence, the selenium webdriver mcp setup allows AI assistants to leverage the power of Selenium Webdriver for web automation, by communicating with a dedicated Selenium MCP server via the Model Context Protocol. This facilitates tasks such as automated web interactions, testing, and data extraction, all controlled by AI.

[图片]

🚀 Overview

A Model Context Protocol (MCP) server for Selenium that provides comprehensive Selenium WebDriver automation tools for AI assistants and applications. This server enables automated web browser interactions, testing, and scraping through a standardized interface.

Built with TypeScript and modern ES modules, it offers type-safe browser automation capabilities through the Model Context Protocol.

✨ Key Features

  • Multi-Browser Support: Chrome, Firefox, Safari, and Edge browser automation
  • Comprehensive Element Interaction: Click, type, hover, drag & drop, file uploads
  • Advanced Navigation: Forward, backward, refresh, window management
  • Wait Strategies: Intelligent waiting for elements and page states
  • Type Safety: Full TypeScript implementation with Zod validation

🤝 Integration

MCP Client Integration

Configure your MCP client to connect to the Selenium server:

Standard Configuration (applicable to Windsurf, Warp, Gemini CLI etc)

json
{  "servers": {    "selenium-mcp": {      "command": "npx",      "args": ["-y", "selenium-webdriver-mcp@latest"]    }  }}

Installation in VS Code

Update your mcp.json in VS Code with below configuration

NOTE: If you're new to MCP servers, follow this link Use MCP servers in VS Code

Example 'stdio' type connection

json
{  "servers": {    "selenium-mcp": {      "command": "npx",      "args": [        "-y",        "selenium-webdriver-mcp@latest"      ],      "type": "stdio"    }  },  "inputs": []}

Example 'http' type connection

json
{  "servers": {    "Selenium": {      "url": "https://smithery.ai/server/@pshivapr/selenium-mcp",      "type": "http"    }  },  "inputs": []}

After installation, the Selenium MCP server will be available for use with your GitHub Copilot agent in VS Code.

To install the Selenium MCP server using the VS Code CLI

bash
# For VS Codecode --add-mcp '{\"name\":\"selenium-mcp\",\"command\": \"npx\",\"args\": [\"selenium-webdriver-mcp@latest\"]}'
bash
# For VS Code Insidersvscode-insiders --add-mcp '{\"name\":\"selenium-mcp\",\"command\": \"npx\",\"args\": [\"selenium-webdriver-mcp@latest\"]}'

To install the package using either npm, or Smithery

Using npm:

bash
npm install -g selenium-webdriver-mcp@latest

Using Smithery

To install Selenium MCP for Claude Desktop automatically via [smithery badge]

bash
npx @smithery/cli install @pshivapr/selenium-mcp --client claude

Claude Desktop Integration

Add to your Claude Desktop configuration:

json
{  "mcpServers": {    "selenium-mcp": {      "command": "npx",      "args": ["-y", "selenium-webdriver-mcp@latest"]    }  }}

Screenshot

[Selenium + Claude]

Prompts

An example prompt to start AI Agent interaction:

Using selenium mcp tools, navigate to <https://parabank.parasoft.com/> click the 'Register' link and signup using dynamic test data and click register. Then generate selenium tests in <YOUR_FAVOURITE_PROGRAMMING_LANGUAGE> using pom, create tests using cucumber features, steps and execute the tests.

Note: For more prompts, look at examples directory of the project

🛠️ MCP Available Tools

Browser Management Tools

ToolDescriptionParameters
browser_openOpen a new browser sessionbrowser, options
browser_navigateNavigate to a URLurl
browser_navigate_backNavigate back in historyNone
browser_navigate_forwardNavigate forward in historyNone
browser_titleGet the current page titleNone
browser_refreshRefresh the current pageNone
browser_get_urlGet the current page URLNone
browser_get_page_sourceGet the current page HTML sourceNone
browser_maximizeMaximize the browser windowNone
browser_resizeResize browser windowwidth, height
browser_closeClose current browser sessionNone

Cookie Management Tools

ToolDescriptionParameters
browser_get_cookiesGet all cookies from the current browser sessionNone
browser_get_cookie_by_nameGet a specific cookie by namecookie (cookie name)
browser_add_cookie_by_nameAdd a new cookie to the browsercookie (cookie name), value
browser_set_cookie_objectSet a cookie object in the browsercookie (cookie object as string)
browser_delete_cookieDelete a specific cookie by namevalue (cookie name to delete)
browser_delete_cookiesDelete all cookies from the current browser sessionNone

Window Management Tools

ToolDescriptionParameters
browser_switch_to_windowSwitch to a different browser window by handlewindowHandle
browser_switch_to_original_windowSwitch back to the original browser windowNone
browser_switch_to_window_by_titleSwitch to a window by its page titletitle
browser_switch_window_by_indexSwitch to a window by its index positionindex
browser_switch_to_window_by_urlSwitch to a window by its URLurl

Element Interaction Tools

ToolDescriptionParameters
browser_find_elementFind an element on the pageby, value, timeout
browser_find_elementsFind multiple elements on the pageby, value, timeout
browser_clickClick on an elementby, value, timeout
browser_typeType text into an elementby, value, text, timeout
browser_get_element_textGet text content of elementby, value, timeout
browser_file_uploadUpload file via input elementby, value, filePath, timeout
browser_clearClear text from an elementby, value, timeout
browser_get_attributeGet element attribute valueby, value, attribute, timeout

Element State Validation Tools

ToolDescriptionParameters
browser_element_is_displayedCheck if an element is visible on the pageby, value, timeout
browser_element_is_enabledCheck if an element is enabled for interactionby, value, timeout
browser_element_is_selectedCheck if an element is selected (checkboxes, radio buttons)by, value, timeout

Frame Management Tools

ToolDescriptionParameters
browser_switch_to_frameSwitch to an iframe elementby, value, timeout
browser_switch_to_parent_frameSwitch to the parent frame (from nested iframe)None
browser_switch_to_default_contentSwitch back to the main page contentNone

Advanced Action Tools

ToolDescriptionParameters
browser_hoverHover over an elementby, value, timeout
browser_double_clickDouble-click on an elementby, value, timeout
browser_right_clickRight-click (context menu)by, value, timeout
browser_drag_and_dropDrag from source to targetby, value, targetBy, targetValue, timeout
browser_wait_for_elementWait for element to appearby, value, timeout
browser_execute_scriptExecute JavaScript codescript, args
browser_screenshotTake a screenshotfilename (optional)
browser_select_dropdown_by_textSelect dropdown option by visible textby, value, text, timeout
browser_select_dropdown_by_valueSelect dropdown option by valueby, value, dropdownValue, timeout
browser_key_pressPress a keyboard key in the browserkey, timeout

Scrolling Tools

ToolDescriptionParameters
browser_scroll_to_elementScroll to bring an element into viewby, value, timeout
browser_scroll_to_topScroll to the top of the pageNone
browser_scroll_to_bottomScroll to the bottom of the pageNone
browser_scroll_to_coordinatesScroll to specific coordinatesx, y
browser_scroll_by_pixelsScroll by specified number of pixelsx, y

Form Interaction Tools

ToolDescriptionParameters
browser_select_checkboxSelect/check a checkboxby, value, timeout
browser_unselect_checkboxUnselect/uncheck a checkboxby, value, timeout
browser_submit_formSubmit a form elementby, value, timeout
browser_focus_elementFocus on a specific elementby, value, timeout
browser_blur_elementRemove focus from a specific elementby, value, timeout

Element Locator Strategies

  • id: Find by element ID
  • css: Find by CSS selector
  • xpath: Find by XPath expression
  • name: Find by name attribute
  • tag: Find by HTML tag name
  • class: Find by CSS class name

📋 Requirements

  • Node.js: Version 18.0.0 or higher
  • Browsers: Chrome, Firefox, Safari, or Edge installed
  • WebDrivers: Automatically managed by selenium-webdriver
  • Operating System: Windows, macOS, or Linux

🚦 Development

Getting Started

Clone the repository
bash
git clone https://github.com/pshivapr/selenium-mcp.gitcd selenium-mcp
Install dependencies
bash
npm install
Build the project
bash
npm run build

Running the Server

Production Mode
bash
npm start
Development Mode (with auto-reload)
bash
npm run dev
Direct Execution
bash
node dist/index.js

Using as CLI Tool

After building, you can use the server as a global command:

bash
npx selenium-webdriver-mcp@latest

📝 License

MIT License - see LICENSE file for details.

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

  1. Fork the repository
  2. Create your feature branch (git checkout -b feature/AmazingFeature)
  3. Commit your changes (git commit -m 'Add some AmazingFeature')
  4. Push to the branch (git push origin feature/AmazingFeature)
  5. Open a Pull Request

Badges/Mentions

MCP Market

Pulse

[MCP Badge]

Built with ❤️ for the Model Context Protocol ecosystem

来源：README.md，提交 bc06d84

工具

0
SourceWeft 尚未在沙箱中启动过该软件包，因此其工具未知。

版本历史

1
  1. v0.4.6最新Sep 16, 2026

评分

已满足适用检查项的 50%

该评分衡量此收录条目的完整程度，依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价，也不表示服务器安全或当前可用：可连通性是带日期的检查结果，评分也不包含安装量或用户评价。

  • 近 30 天内可连通25 分

    请保持端点能够响应 MCP 握手；SourceWeft 每 30 天探测一次。

  • 已声明认证方式10 分

    请在注册中心条目中声明凭据请求头，或保持端点能够响应探测，以便判定其认证方式。

  • 已列出工具15 分

    请让服务器在无凭据时响应 tools/list，或发布可在沙箱中无凭据启动的软件包。

  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分