// the take
A WeChat automation stack split into a Windows server that drives the desktop WeChat client via UI Automation, and a Django+Celery client that handles scheduling, admin, and alerting. It's for people who want to batch-send or schedule WeChat messages/files, scrape chat logs for keyword alerts, or auto-claim red packets, and are willing to dedicate a Windows box to it.
The server/client split is a sound design: the fragile UI-automation part stays isolated on the one machine that actually needs it, while scheduling, Postgres storage, and email alerting live in a normal Docker-deployable Django app talking to it over HTTP. A single global request queue with a mutex serializes every WeChat action, which is the right call for UI automation - it avoids the classic failure mode of two concurrent automations fighting over the same window and sending a message to the wrong contact. The auto red-packet/transfer claiming explicitly skips group chats to limit the blast radius of unattended fund handling, and request logging captures queued/running/success/failed per call plus client IP, which makes a flaky run debuggable after the fact instead of just silently missing a send.
The V3 server ships with zero API authentication and the README just tells you to keep it off the public internet - fine advice, but the tool can send messages and auto-claim money, so a misconfigured firewall or a LAN with other untrusted devices is a real exposure, not a theoretical one. It's built entirely on Windows UI Automation against one pinned WeChat build (4.1.8, zh-CN only); WeChat ships frequent updates and this kind of thing breaks the moment the UI tree shifts, with no fallback besides re-running a 'restart Narrator, relogin' ritual. Automating red packet and transfer claims via UI scripting with no transactional confirmation is risky by the project's own admission - the docs explicitly warn not to treat it as an accounting source of truth. Operationally it needs a dedicated Windows machine that stays logged into WeChat, unlocked, and never sleeps, which is a heavier and more fragile ask than most people scheduling a few messages are expecting, and the single serial queue means it never scales past one WeChat session no matter how many clients talk to it.
We read the README, the directory tree and a key source file in November 2026. Nobody pays to be listed. How we judge →