Weixin MCP - WeChat Article Reader

A minimalist MCP that allows large models to read WeChat public account articles.

Core Features

• 🎭 Browser Simulation: Uses Playwright to fully simulate a browser environment

• 📝 Content Extraction: Automatically extracts title, author, publication time, and body content

• ⚡ Concise Implementation: Core functionality implemented with minimal code

Related MCP server: WeChat Article Parser

Workflow

1. User sends URL and requirements to the large model

2. Large model calls the MCP tool

3. MCP retrieves article content and sends it to the large model

4. Large model outputs natural language based on the article content

Tech Stack

• Python 3.10+

• fastmcp - MCP framework

• url-md (Rust single binary) - Anti-scraping + Markdown extraction in one step

• pyyaml - frontmatter parsing

v0.3.0 Upgrade Note: The scraping layer has been simplified from 4 agent-browser subprocesses + BeautifulSoup parsing to a single call to url-md md <url>. url-md internally handles anti-scraping / WeChat body extraction / Markdown conversion / frontmatter generation. Dependencies are reduced, and the content field is upgraded to Markdown (preserving image references / lists / heading levels). There are zero changes to the MCP protocol interface, so existing Claude/Cursor configurations do not need adjustment.

v0.2.0 Upgrade Note: The original Playwright solution was broken by WeChat's strengthened anti-scraping (issue #3). The scraping layer was delegated to agent-browser (Apache-2.0 open-source Rust project). As of v0.3.0, it has migrated to url-md.

Quick Start

1. Install url-md (Required from v0.3.0 onwards)

# macOS / Linux
curl -fsSL https://raw.githubusercontent.com/Bwkyd/url-md/main/install.sh | bash

# Windows (PowerShell)
irm https://raw.githubusercontent.com/Bwkyd/url-md/main/install.ps1 | iex

# 验证
url-md --version

6 seconds from zero to ready. 7 MB single binary, no external dependencies like Chrome required (WeChat permanent links use the fast reqwest path).

2. Install Python Dependencies

pip install -r requirements.txt

3. Configuration

{
  "mcpServers": {
    "weixin-reader": {
      "command": "python",
      "args": [
        "C:/Users/你的用户名/Desktop/wx-mcp/wx-mcp-server/src/server.py"
      ]
    }
  }
}

Note: Please replace the path with your actual project path.

Usage Example

Use directly in Claude:

请帮我总结这篇文章：https://mp.weixin.qq.com/s/nEJhdxGea-KLZA_IGw9R5A

Claude will automatically call the read_weixin_article tool to retrieve the article content and analyze it.

Function Description

read_weixin_article(url: str)

Read the content of a WeChat public account article.

Parameters:

• url: WeChat article URL, format: https://mp.weixin.qq.com/s/xxx

Returns:

{
  "success": true,
  "title": "文章标题",
  "author": "作者名",
  "publish_time": "2025-11-05",
  "content": "# 文章正文\n\n![封面](https://...)\n\n段落内容...",
  "cover_url": "https://mmbiz.qpic.cn/.../wx_fmt=jpeg",
  "error": null
}

As of v0.3.0, the content field is in Markdown format (it was plain text in v0.2.0 and earlier). If downstream agents/prompts are already handling it as plain text, you can strip the Markdown syntax before returning, or keep the Markdown to allow the LLM to understand it better natively.

Notes

• ⚠️ For personal learning and research only

• ⚠️ Comply with the WeChat Public Platform Service Agreement

• ⚠️ High-frequency scraping is not recommended (suggested interval > 2 seconds)

• ⚠️ Do not use for commercial purposes