The data half is the useful half: dashboard, per-note performance and follower figures come back with Chinese headers already in place, so an assistant can compare notes against each other instead of transcribing a screenshot.
A Xiaohongshu creator toolkit that drives the site through a signed-in Chrome session. It publishes notes on your behalf and collects creator-centre data on a schedule, in Chinese-headed CSV that a model can read directly.
- `smart_publish_note` — publish an image or video note with title, content, images, videos, tags and topics; images can be local paths or remote URLs, and topic tags are added automatically
- `check_task_status` and `get_task_result` — follow a publish job to completion and collect its final result
- `login_xiaohongshu` — sign in from the conversation, with `force_relogin` and `quick_mode`, so no separate terminal step is needed
- `get_creator_data_analysis` — the creator data set assembled for analysis: account overview such as followers and likes, per-note performance such as views and likes, and follower growth
- `test_connection` — a plain connection check
- Scheduled collection driven by cron expressions, stored locally as CSV
Google Chrome plus a ChromeDriver whose version matches it exactly — a mismatch is the most common cause of failure. `webdriver-manager` can manage that automatically. A signed-in Xiaohongshu cookie is required and is what authorizes everything. The model on the other end must be vision-capable; text-only models will not work. Run it with `./xhs` on macOS and Linux or `xhs.bat` on Windows, or from source with `uv sync`; configuration lives in a `.env` file copied from `env_example`, with `CHROME_PATH` pointing at your browser. `ENABLE_REMOTE_BROWSER`, `REMOTE_BROWSER_HOST` and `REMOTE_BROWSER_PORT` connect to an already-running Chrome instead, which is how the Docker and remote-deployment paths work. Note the author has stopped maintaining the project and has said there are no plans to resume; the automation and signing work is there to fork.
Build from source — clone the repository and build it, then point your client at the binary