Through a locally deployed wechat-download-api service, batch download all historical articles (including full content) of any WeChat official account.
Trigger conditions
- The user asks to download/crawl WeChat official account articles
- The user asks to search for an official account and get its articles
- The user says "download the articles of XX official account"
- The user mentions wechat-download / 公众号下载 / 微信文章抓取
Architecture
本 skill (调用脚本) → wechat-download-api (本地 HTTP 服务 :8765) → 微信公众平台 API
- wechat-download-api: third-party open-source service, uses Chrome TLS fingerprint (curl_cffi) to simulate a browser, and calls the API through WeChat official platform backend credentials
- This skill's script: encapsulates the API's search, list, and download functions, with adaptive rate limiting and resumable downloads
Usage
After the user installs this skill, they only need to say in the agent "download the official account articles of XXX". Except for QR code login, do not ask the user to manually deploy the underlying service; the agent should prioritize automatically running the scripts in this directory to complete environment preparation.
1. Start the service
Run from the skill directory:
cd ~/.claude/skills/wechat-download
bash scripts/start_wechat_api.sh
# 默认端口 8765,可指定: bash scripts/start_wechat_api.sh 9876
The script will automatically discover ./wechat-download-api、../wechat-download-api, and also supports locating it through the WECHAT_API_REPO environment variable. If it cannot be found, it will automatically clone wechat-download-api to the current skill directory and install the underlying service dependencies.
Port note: this skill uses
8765by default instead of upstream's5000, because on macOS5000is often occupied by system services. If you started the service with upstream'sstart.sh(port5000), please add before the commandWECHAT_API_URL=http://localhost:5000。
The script will automatically:
- Find or download the wechat-download-api repository
- Create
.env(if it does not exist) - Install missing Python dependencies
- Start the service in the background
2. QR code login
Prerequisite: the user needs to have a WeChat official account (subscription account/service account are both fine).
- Open in a browser
http://localhost:8765/login.html - Scan the QR code with the WeChat of the official account administrator
- After successful login, the credentials are automatically saved to
.env, valid for about 4 days
If the user does not have an official account, see "Alternative solution" below.
Operation flow
Step 1: Check/start the service
If the service has not been started yet, run first:
bash scripts/start_wechat_api.sh
Then check the service status:
python3 scripts/download_wechat.py status
Step 2: Check login
curl -s http://localhost:8765/api/admin/status
If isExpired: true → scan the QR code again to log in.
Step 3: Search for the official account
python3 scripts/download_wechat.py search "公众号名称"
Get from the output fakeid。
Step 4: Download articles
python3 scripts/download_wechat.py download "fakeid" 200 ./output_dir
Features:
- Automatically paginate to get the full article list
- Download the full content article by article, save as
.txtfiles - Resumable downloads: already downloaded files are automatically skipped
- Adaptive rate limiting: automatically waits when Rate limited is detected
- Failure records are saved to
failures.json
Other commands
python3 scripts/download_wechat.py list "fakeid" 0 50 # 查看文章列表
python3 scripts/download_wechat.py rss "fakeid" # 获取 RSS 地址
Common errors and solutions
Error 1: Port is occupied
ERROR: [Errno 48] address already in use
Solution: switch to another port:bash scripts/start_wechat_api.sh 9876
Error 2: Article URL field name (critical)
url = a.get("url", "") # ❌ 错误!
url = a.get("link", "") # ✓ 正确
The article list returned by the API uses the link field, not url。
Error 3: Rate Limited
Rate limited: 请求过于频繁,请X秒后重试
Solution: The script has built-in adaptive backoff. If it is still triggered frequently, increase MIN_DELAY(default 3.2 seconds).
Error 4: WeChat security verification triggered
触发微信安全验证。
Solution:
- Reduce request frequency
- Open the prompted article URL in a browser to complete the human verification
- Wait 30 minutes and retry
- If triggered frequently, configure
.envinPROXY_URLS(SOCKS5 proxy pool)
Error 5: Unable to get article content
无法获取文章内容。可能原因:文章被删除、访问受限或需要验证。
Cause: The article has been deleted / is an advertisement / is an image share / WeChat risk control. Just skip it; this usually accounts for 30-50%.
Error 6: Login expired
Cause: WeChat credentials are valid for about 4 days.
Solution: Revisit http://localhost:8765/login.html to scan the QR code.
Error 7: curl_cffi compilation failed
ModuleNotFoundError: No module named 'curl_cffi'
Solution: A working C compiler is required. macOS requires installing Xcode CLT (xcode-select --install), Linux requires build-essential and libcurl4-openssl-dev. Precompiled wheels on PyPI cover mainstream platforms.
Alternative: Download without an official account
If the user does not have a WeChat official account and cannot log in, the following methods can be used:
- Search engine reverse lookup: Search
"公众号名" mp.weixin.qq.com/sto get the article link, and directly visit mp.weixin.qq.com to extract the content in#js_contentcontent in - Reposting platforms: Search for reposted articles on platforms such as guancha.cn (Guancha), 36kr, Sina Finance, etc.
- Mirror sites: Platforms such as jintiankansha.me have article lists (but usually require VIP to view the full text)
Reference
- wechat-download-api — underlying API service, AGPL 3.0
- Article URL format description:
https://mp.weixin.qq.com/s/{short_id}or a complete link with__biz/mid/idx/snparameters