Markdown MCP Server

MCP浏览器工具

提供方 vishwajeetdabholkar · 包名 playwright

浏览器工具 stdio · 本地进程 会话登录 社区 / 第三方 Claude DesktopClaude
查看接入方式

MCP服务器通过Playwright从网页中提取干净的Markdown内容,适用于文档整理、内容分析和AI集成等场景。

一句话选型结论

按资料字段整理

Markdown MCP Server 属于浏览器工具类 MCP 服务器,stdio · 本地进程,会话登录,可直接接入 Claude Desktop、Claude,有可直接执行的安装命令。接入前先确认权限范围;涉及写入、删除或付款的操作,建议先在测试环境验证。

分类浏览器工具 传输方式stdio · 本地进程 鉴权会话登录 部署方式本地部署
GitHub 星标 1 仓库热度,无公开仓库时显示未标注
工具数 1 暴露给模型的工具数量
资源 / 提示词 0 / 0 资源与提示词模板数量
维护方 社区 / 第三方 核验等级 仅资料核对

工具与能力

浏览器工具
数量来自来源目录的标注,具体工具清单以官方文档为准
工具数量1暴露给模型可调用的工具个数
资源数量0可读取的资源条目
提示词数量0内置提示词模板数量

能力标签

支持客户端

适用场景

网页抓取JavaScript本地部署内容提取Markdown转换自动化工具AI集成

安装与配置

stdio · 本地进程 会话登录
客户端在本机拉起子进程,数据不出本机
npx playwright install chromium

这条命令只是入口,具体启动参数请按官方文档补齐,不要直接照搬到别的服务器。

  1. 1
    准备运行环境

    stdio 服务在本地起进程,需要先装好 Node.js,再按安装命令拉取依赖。

  2. 2
    安装服务

    复制上面的安装命令执行;命令里带占位符的位置,换成自己的密钥或路径。

  3. 3
    写入客户端配置

    在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,让新配置生效。

  4. 4
    验证是否可用

    连接成功后客户端会列出约 1 个工具,随便调一个只读工具确认能返回结果。

典型使用场景

3 类
按分类与标签整理的常见用法,具体以官方文档为准
自动操作网页

打开页面、点击、填表,把浏览器动作交给模型执行。

抓取结构化数据

把页面内容整理成表格或 JSON,省掉手工复制粘贴。

关键流程回归

按步骤检查页面是否正常,回归时少漏环节。

选择建议

选之前先看这几条
按条目字段给出的判断,不替你决定用不用

值得选的地方

  • 资料标注 1 个工具,职责相对集中,接入后不容易误操作。
  • 有现成的安装命令,不用自己找入口,照着跑就能起服务。
  • 已适配 Claude Desktop、Claude,换客户端不用重新研究配置。

先想清楚的地方

  • stdio 方式要求本地装好运行时(如 Node.js),换机器要重新装一遍。
  • 使用前要准备好会话登录对应的凭据,并按最小权限配置。
  • 版本更新与行为细节以官方仓库为准。

更适合

  • 需要在浏览器工具场景里让模型直接调用外部能力的人。
  • 已经在用 Claude Desktop、Claude 的用户,接入路径最短。
  • 想先快速试一下、能接受在本地或自有环境跑服务的人。

不太适合

  • 完全不想在本地装环境的人。
  • 不方便管理密钥或第三方授权的人。
  • 需要厂商级 SLA 和技术支持的生产环境。

同类对比

6 个同类
同分类里挑最接近的几条,字段以各自详情页为准
MCP 服务器工具数传输方式部署方式安装命令最后核验
Markdown MCP Server当前 1 stdio · 本地进程 本地部署 有 2026-05-17
Playwright MCP Server 21 stdio · 本地进程 本地部署 有 2026-05-18
Fetch 9a2682de 1 SSE · 服务端推送 支持远程部署 有 2026-05-18
Browser 738d52bc 未标注 stdio · 本地进程 本地部署 有 2026-05-17
MCP 40600ece 未标注 stdio · 本地进程 本地部署 有 2026-05-17
Skyvern MCP 未标注 stdio · 本地进程 本地部署 有 2026-05-17

服务器档案

浏览器工具
字段来自项目仓库与公开索引,中文说明由本站标注
提供方vishwajeetdabholkar上游标注的维护方 vishwajeetdabholkar
分类浏览器工具本站按上游分类展示
连接方式stdio · 本地进程客户端在本机拉起子进程,数据不出本机
鉴权方式会话登录远程服务建议用最小权限凭据
维护方社区 / 第三方核验等级 仅资料核对
包名 / 运行时playwright · Node.js本地安装时用到的包标识与运行时
部署类型本地部署上游给出的部署形态
最近核验2026-05-17上游最后核验时间

详细介绍

13.2 KB
来自项目仓库与上游目录的原文,未做删改

Markdown MCP Server

A Model Context Protocol (MCP) server that extracts clean markdown content from web pages using Playwright. This server provides a get_page_markdown tool that can extract the main content from any URL while filtering out navigation, headers, footers, and other non-content elements.

Features

  • 🎯 Smart Content Extraction: Automatically identifies and extracts main content from web pages
  • 🧹 Clean Output: Filters out navigation, headers, footers, sidebars, and advertisements
  • 🎨 Rich Markdown: Preserves formatting including headings, bold, italic, code blocks, lists, and tables
  • 🖼️ Image Support: Optionally includes image references in markdown
  • 🔗 Link Support: Optionally includes hyperlinks in markdown
  • ⚡ Fast & Reliable: Uses Playwright for robust web scraping
  • 🔄 Dynamic Content: Handles JavaScript-heavy sites and dynamic content loading
  • 🛡️ Error Handling: Robust error handling with fallback extraction methods

Installation

  1. Clone or download this repository:bash git clone <repository-url> cd markdown-mcp

  2. Install dependencies:bash npm install

  3. Install Playwright browsers:bash npx playwright install chromium

  4. Make the script executable (optional):bash chmod +x markdown-mcp.js

Usage

As an MCP Server

Start the server:```bash
node markdown-mcp.js


The server provides one tool: `get_page_markdown`

#### Tool Parameters

- `url` (required): The URL to extract markdown from
- `includeImages` (optional, default: true): Whether to include image references in markdown
- `includeLinks` (optional, default: true): Whether to include hyperlinks in markdown
- `waitForSelector` (optional): CSS selector to wait for before extracting content (useful for dynamic content)
- `timeout` (optional, default: 30000): Navigation timeout in milliseconds

#### Example Usage```json
{
  "name": "get_page_markdown",
  "arguments": {
    "url": "https://docs.confluent.io/cloud/current/flink/operate-and-deploy/monitor-statements.html",
    "includeImages": true,
    "includeLinks": true,
    "timeout": 30000
  }
}

Advanced Usage Examples

Extract content from a specific section:```json
{
"name": "get_page_markdown",
"arguments": {
"url": "https://example.com/article",
"waitForSelector": ".main-content",
"includeImages": false,
"includeLinks": true
}
}


**Extract content with custom timeout:**```json
{
  "name": "get_page_markdown",
  "arguments": {
    "url": "https://slow-loading-site.com",
    "timeout": 60000
  }
}

File Structure

This project includes two MCP server files optimized for different clients:

  • markdown-mcp.js - Optimized for Claude Desktop
  • markdown-mcp-gemini.js - Optimized for Gemini CLI

Both files provide the same get_page_markdown tool but are configured differently for optimal performance with each client.

Adding to AI Clients

This MCP server can be used with multiple AI clients that support the Model Context Protocol. Below are instructions for the most popular clients.

Claude Desktop Integration

To use this MCP server with Claude Desktop, you need to add it to your Claude Desktop configuration file.

Step 1: Locate Claude Desktop Configuration

macOS:
- Configuration file: ~/Library/Application Support/Claude/claude_desktop_config.json

Windows:
- Configuration file: %APPDATA%\Claude\claude_desktop_config.json

Linux:
- Configuration file: ~/.config/claude/claude_desktop_config.json

Step 2: Edit Configuration File

  1. Open the configuration file in a text editor
  2. Add the markdown-mcp server to the mcpServers section
  3. Update the path to point to your markdown-mcp.js file

Step 3: Configuration Examples

macOS Configuration```json

{
"mcpServers": {
"markdown-mcp": {
"command": "node",
"args": ["/Users/yourusername/path/to/markdown-mcp/markdown-mcp.js"],
"env": {}
}
}
}


#### Windows Configuration```json
{
  "mcpServers": {
    "markdown-mcp": {
      "command": "node",
      "args": ["C:\\Users\\YourUsername\\path\\to\\markdown-mcp\\markdown-mcp.js"],
      "env": {}
    }
  }
}

Linux Configuration```json

{
"mcpServers": {
"markdown-mcp": {
"command": "node",
"args": ["/home/yourusername/path/to/markdown-mcp/markdown-mcp.js"],
"env": {}
}
}
}


### Step 4: Restart Claude Desktop

After updating the configuration file, restart Claude Desktop for the changes to take effect.

### Step 5: Verify Installation

1. Open Claude Desktop
2. Start a new conversation
3. Try asking Claude to extract content from a webpage using the markdown-mcp tool
4. Example: "Use markdown-mcp to extract content from https://example.com"

### Troubleshooting

**If the MCP server doesn't work:**

1. **Check the file path** - Make sure the path to `markdown-mcp.js` is correct and the file exists
2. **Verify Node.js** - Ensure Node.js is installed and accessible from the command line
3. **Check permissions** - Make sure the script has execute permissions
4. **Test manually** - Try running `node markdown-mcp.js` in the terminal to see if there are any errors
5. **Check Claude Desktop logs** - Look for error messages in Claude Desktop's developer console

**Common Issues:**

- **Path not found**: Double-check the file path in the configuration
- **Node.js not found**: Make sure Node.js is installed and in your PATH
- **Permission denied**: Run `chmod +x markdown-mcp.js` to make the script executable
- **Dependencies missing**: Run `npm install` in the markdown-mcp directory

## Gemini CLI Integration

To use this MCP server with Gemini CLI, follow these steps:

### Step 1: Install Gemini CLI

If you haven't already installed Gemini CLI:```bash
npm install -g @google/gemini-cli

Verify the installation:```bash
gemini --version


### Step 2: Add MCP Server to Gemini CLI

Add your markdown-mcp server to Gemini CLI:```bash
gemini mcp add markdown-mcp /Users/yourusername/path/to/markdown-mcp/markdown-mcp-gemini.js

Important: Replace /Users/yourusername/path/to/markdown-mcp/markdown-mcp-gemini.js with the actual path to your markdown-mcp-gemini.js file.

Step 3: Verify Integration

List all configured MCP servers to verify the integration:```bash
gemini mcp list


You should see `markdown-mcp` listed among the servers.

### Step 4: Test the Integration

Test the markdown-mcp server with Gemini CLI:```bash
# Example: Extract content from a webpage
gemini "Use the markdown-mcp tool to extract content from https://example.com"

Or you can use the tool directly:```bash

If the tool is exposed as a command

gemini get_page_markdown "https://example.com"


### Step 5: Complete Example - Extract and Save Markdown

Here's a complete example that extracts markdown content and saves it to a file:```bash
# Extract content from a webpage and save to result.md
gemini "Use get_page_markdown to extract content from https://www.confluent.io/blog/event-driven-flink-agents-enterprise-ai/ and save the response as result.md"

This command will:
1. Use the get_page_markdown tool to extract clean markdown content from the Confluent blog post
2. Save the extracted markdown content to a file named result.md in your current directory
3. Provide you with a clean, readable markdown version of the webpage content

Additional Examples:```bash

Extract content from documentation and save with custom filename

gemini "Use get_page_markdown to extract content from https://docs.confluent.io/cloud/current/flink/operate-and-deploy/monitor-statements.html and save it as flink-docs.md"

Extract content from a GitHub repository README

gemini "Use get_page_markdown to extract content from https://github.com/microsoft/vscode and save as vscode-readme.md"

Extract content with specific options

gemini "Use get_page_markdown with includeImages=false to extract content from https://example.com and save as clean-content.md"


### Gemini CLI Troubleshooting

**If the MCP server doesn't work with Gemini CLI:**

1. **Check the file path** - Ensure the path to `markdown-mcp-gemini.js` is correct and absolute
2. **Verify Node.js** - Make sure Node.js is accessible from the command line
3. **Check permissions** - Ensure the script has execute permissions (`chmod +x markdown-mcp-gemini.js`)
4. **Test the server manually** - Run `node markdown-mcp-gemini.js` to check for errors
5. **Check Gemini CLI logs** - Look for error messages in the Gemini CLI output

**Common Gemini CLI Issues:**

- **Path not found**: Use absolute paths when adding the MCP server
- **Permission denied**: Run `chmod +x markdown-mcp-gemini.js` to make the script executable
- **Node.js not found**: Ensure Node.js is installed and in your PATH
- **Server not responding**: Check if the server starts correctly with `node markdown-mcp-gemini.js`

## Using with Multiple AI Clients

You can use the same markdown-mcp server with multiple AI clients simultaneously. The MCP server is designed to handle multiple concurrent requests efficiently.

### Benefits of Multi-Client Setup

- **Flexibility**: Use the same tool with different AI models
- **Efficiency**: Share the same server instance across clients
- **Consistency**: Get the same extraction quality regardless of the AI client
- **Resource optimization**: No need to run multiple server instances

### Setup for Multiple Clients

1. **Set up Claude Desktop** using `markdown-mcp.js` (as described above)
2. **Set up Gemini CLI** using `markdown-mcp-gemini.js` (as described above)
3. **Both clients can use their respective server files** - optimized for each client

### Usage Examples

**With Claude Desktop:**```
Use markdown-mcp to extract content from https://docs.confluent.io/cloud/current/flink/operate-and-deploy/monitor-statements.html

With Gemini CLI:```bash

Extract and save to file

gemini "Use get_page_markdown to extract content from https://docs.confluent.io/cloud/current/flink/operate-and-deploy/monitor-statements.html and save as result.md"

Or just extract without saving

gemini "Use get_page_markdown to extract content from https://docs.confluent.io/cloud/current/flink/operate-and-deploy/monitor-statements.html"


### Performance Considerations

- The server handles multiple concurrent requests efficiently
- Each request uses a fresh browser context for security
- Memory usage scales with the number of concurrent requests
- Typical response time: 5-15 seconds per request

## Testing

The server has been tested and verified to work correctly with various websites including:

- ✅ **Documentation sites** (Confluent, GitHub, etc.)
- ✅ **News articles** and blog posts
- ✅ **Technical documentation** with code examples
- ✅ **E-commerce pages** and product descriptions
- ✅ **JavaScript-heavy sites** with dynamic content

### Tested Features

- ✅ Extracts headings, paragraphs, and text content
- ✅ Preserves bold and italic formatting
- ✅ Handles code blocks and inline code
- ✅ Processes lists (ordered and unordered)
- ✅ Extracts tables with proper formatting
- ✅ Filters out navigation and footer content
- ✅ Handles images and links (when enabled)
- ✅ Responds to MCP protocol requests
- ✅ Works with dynamic content and JavaScript-heavy sites

### Manual Testing

You can test the server manually by running:```bash
# Test with a simple URL
node -e "
const { spawn } = require('child_process');
const server = spawn('node', ['markdown-mcp.js'], { stdio: ['pipe', 'pipe', 'pipe'] });
const request = {
  jsonrpc: '2.0',
  id: 1,
  method: 'tools/call',
  params: {
    name: 'get_page_markdown',
    arguments: { url: 'https://example.com' }
  }
};
server.stdin.write(JSON.stringify(request) + '\n');
setTimeout(() => {
  server.kill();
  console.log('Test completed');
}, 10000);
"

Supported Websites

This MCP server works well with:

  • Documentation sites: Confluent, GitHub, GitLab, etc.
  • News and blogs: Most major news sites and blogs
  • Technical content: Stack Overflow, Medium, Dev.to
  • E-commerce: Product pages and descriptions
  • Academic content: Research papers and articles
  • Social media: Twitter threads, LinkedIn articles

Performance

  • Typical extraction time: 5-15 seconds depending on page complexity
  • Memory usage: ~50-100MB per extraction
  • Supported content size: Up to several MB of text content
  • Concurrent requests: Handles multiple requests efficiently

Requirements

  • Node.js: Version 18 or higher
  • Playwright: Chromium browser (installed automatically)
  • Memory: At least 512MB available RAM
  • Disk space: ~200MB for Playwright browser

Security Considerations

  • The server runs in headless mode for security
  • No cookies or persistent data is stored
  • Each request uses a fresh browser context
  • Network requests are limited by timeout settings
  • No sensitive data is logged or stored

Contributing

  1. Fork the repository
  2. Create a feature branch
  3. Make your changes
  4. Test thoroughly
  5. Submit a pull request

Support

If you encounter issues:

  1. Check the troubleshooting section above
  2. Verify all requirements are met
  3. Test with a simple URL first
  4. Check Claude Desktop logs for error messages
  5. Open an issue with detailed error information

常见问题

8 问
答案由本站依据条目字段整理,不引用上游原文
Markdown MCP Server 怎么安装?

在客户端按下面的安装命令接入:npx playwright install chromium。对应的包名是 playwright。装完在客户端配置里引用同一个包或端点。

Markdown MCP Server 怎么在 Claude Desktop、Claude 里配置?

它的传输方式是stdio · 本地进程,在客户端的 MCP 配置里填好启动命令和参数,保存后完全退出客户端再打开,配置才会生效。

Markdown MCP Server 收费吗?

本站不代收费用,也没有标注具体价格。是否收费取决于服务本身和背后的接口额度,正式使用前按官方说明确认。

Markdown MCP Server 需要 API Key 吗?

资料标注需要会话登录认证,使用前准备好对应凭据,并按最小权限配置。

Markdown MCP Server 支持哪些客户端?

资料标注支持 Claude Desktop、Claude;其他客户端只要能配置 MCP Server,也能用同样的命令或地址接入。

Markdown MCP Server 能提供哪些能力?

资料显示约 1 个工具;能力标签包括 浏览器自动化、内容提取、JavaScript、Claude、Markdown转换。具体工具清单以官方文档为准。

Markdown MCP Server 需要什么运行环境?

运行时为 Node.js;属于本地进程,需要在本机装好运行环境。

Markdown MCP Server 是本地还是远程?

它是本地进程(stdio),由客户端在本机拉起。传输方式为stdio · 本地进程。

热门专题

按标签挑服务器
全部专题 →

服务器信息以作者仓库与官方文档为准;分类、传输方式与鉴权方式的中文说明为中文整理,最近核验 2026-05-17。