CN Scraper MCP — Claude Code MCP server: search Chinese platforms
Gives your agent tools to search Taobao, JD, Xiaohongshu, Zhihu, Weibo, Bilibili and more, with cookies kept on your machine.
claude mcp add cn-scraper-mcp -- uvx cn-scraper-mcpWhat CN Scraper MCP does
Agents can search the web, but Chinese platforms sit behind logins, browser fingerprints or platform-specific parameters. This server packages those workarounds into MCP tools: taobao_search, jd_search, xiaohongshu_search, zhihu_search, weibo_search, bilibili_search, douban_search, dianping_search, zsxq_topics and others. Some platforms run over plain HTTP, others need a local Chrome.
Install with pip install cn-scraper-mcp, then start it with cn-scraper-mcp or python -m cn_scraper_mcp.server. Python 3.11+ is required. The easiest login path is guided_login(platform="weibo"): it opens the official page, you sign in yourself, and the tool reads the cookies through CDP into ~/.cn-scraper-cookies/. JD keeps its session in a local Chrome Profile. A Docker image ships with Chromium inside, and an HTTP transport mode covers remote setups.
Know the limits before you wire it up. Xiaohongshu only accepts residential IPs — datacenter addresses get blocked, so a local Chrome is mandatory there. Douyin is marked experimental: search needs a login and a manual captcha. Pinduoduo returns results only on the first search per browser session and then reports a system error, so the author advises against it. In HTTP mode, don't expose the port publicly: the tools reach your local cookies and profile.
What it helps with
- pull Taobao or JD listings with prices and sellers
- collect Xiaohongshu notes and comments
- get Zhihu answers and hot topics
- export Weibo posts and trending searches
- read Bilibili videos and comments
Example prompts for the agent
- Search Taobao for kids' study desks and show prices and shops
- Collect Xiaohongshu notes about skincare
- Show today's Zhihu hot list
How to install CN Scraper MCP
Platform CLI
claude mcp add cn-scraper-mcp -- uvx cn-scraper-mcpuvx runs the command named after the package; if the README names another one, use uvx --from cn-scraper-mcp <command>
Platform CLI
codex mcp add cn-scraper-mcp -- uvx cn-scraper-mcpuvx runs the command named after the package; if the README names another one, use uvx --from cn-scraper-mcp <command>
Config file
{
"mcpServers": {
"cn-scraper-mcp": {
"command": "uvx",
"args": [
"cn-scraper-mcp"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from cn-scraper-mcp <command>
Config file
{
"servers": {
"cn-scraper-mcp": {
"type": "stdio",
"command": "uvx",
"args": [
"cn-scraper-mcp"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from cn-scraper-mcp <command>
Platform CLI
gemini mcp add cn-scraper-mcp uvx cn-scraper-mcpuvx runs the command named after the package; if the README names another one, use uvx --from cn-scraper-mcp <command>
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about CN Scraper MCP
How do I install CN Scraper MCP in Claude Code?
Run claude mcp add cn-scraper-mcp -- uvx cn-scraper-mcp. That registers the server in Claude through uvx.
What does CN Scraper MCP do?
It gives an agent search and read tools for Chinese platforms: Taobao, JD, Xiaohongshu, Zhihu, Weibo, Bilibili, Douban, Dianping and Zsxq. Cookies and the browser profile stay local instead of going to a third-party server.
Do I need a browser for it to work?
For some platforms, yes. JD, Xiaohongshu, Pinduoduo and Douyin go through a local Chrome over CDP. Taobao, Zhihu, Weibo, Bilibili, Douban and Dianping work without a browser.
