crawl_wechat_article
Returns crawl instructions for one WeChat Official Account article, which the agent then executes through playwright-mcp; it does not fetch the page itself.
LABSCO SUMMARY
The images are the hard part of saving a WeChat article, and that is what it solves.
Copying the text out is easy; keeping the pictures is not, because WeChat serves them from a separate domain that rejects requests without the right Referer and encodes the format in a query parameter — this handles both and rewrites the Markdown to point at the local files, so the saved article still reads a year later. Check how it runs in your setup before planning around it: the tool describes itself as returning crawl instructions for the agent to execute through playwright-mcp, while the project also documents a mode that downloads directly.
What it is
A single-purpose crawler for WeChat articles, built for Cursor. It handles the parts that make mp.weixin.qq.com awkward: the mmbiz.qpic.cn image domain, the Referer and User-Agent headers those images require, and the wx_fmt parameter that decides the file type.
What you get
- crawl_wechat_article takes the article URL and returns the content, with output_format choosing markdown, json or html
- save_images downloads the article's images locally and rewrites the document's paths to them, so the Markdown reads offline
- clean_content strips the page furniture, and a timeout bounds a slow fetch
- Strategy picks between fast, basic and conservative — the project's own trade-off between speed and reliability on a flaky page or a complex article
- Image downloading runs three at a time with retries, rather than serially or all at once
- The tool description states it returns crawl instructions for the agent to execute through playwright-mcp; the project also documents an automatic mode that performs the download itself
- The documentation additionally describes crawl_wechat_batch for a list of URLs with a concurrency limit and inter-request delay, and crawl_get_status for session progress
Requirements
No account and no key. Node.js 18 or higher — it uses the built-in fetch — run as npx -y crawl-mcp-server. The instruction path expects playwright-mcp to be available to the same agent, since that is what performs the page fetch. Documentation and tool descriptions are in Chinese. MIT.
Setup effort
One command — npx -y crawl-mcp-server@1.1.0
USE CASES
01
Keeping an article that might disappear
Save it as Markdown with its images downloaded and the paths rewritten, so the local copy is complete.
02
Getting an article's text into a document you are writing
Pull it with the page furniture stripped, in the format that suits where it is going.
03
Fetching an article that keeps timing out
Switch to the conservative strategy and raise the timeout rather than retrying the fast path.
04
Building a local archive of a public account you follow
Save each article as Markdown with its images, so the set stays readable without a network connection.
1 tool
Counted once with only startup setup. Other versions, modes or optional credentials may expose a different list.
BEFORE YOU INSTALL
What to know first
Tools you get
1 tool when we connected; the project documents 3. The documentation also describes crawl_wechat_batch and crawl_get_status, with no documented flag to enable them; the pinned crawl-mcp-server@1.1.0 package likely predates them.
Requires
- npx on PATH
- Node.js 18+
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
Server setup
TERMINAL
npx -y crawl-mcp-server@1.1.0MCP Clients
Cursor
CURSOR · JSON
{
"mcpServers": {
"crawl-mcp": {
"command": "npx",
"args": [
"-y",
"crawl-mcp-server@1.1.0"
],
"env": {
"NODE_ENV": "production"
}
}
}
}
