Xiaohongshu (RedNote), Douyin and Weibo API for market research

1. Choose keywords

   Search the words your customers use: 露营 (camping), 咖啡 (coffee), a product category or a brand name. Douyin and Weibo also take English queries and brand names.

2. Run the search

   Send the keyword to POST /v1/tools/{id}/run with a max_results cap. The run holds its maximum cost and returns the results as JSON.

3. Read, translate, summarize

   Results come back in their original language. Send them to a model with the same key to translate, summarize or tag them in English.

Xiaohongshu search runs one operation per call. search_notes returns notes for a keyword; this run asks for the 20 most-liked camping notes:

curl https://lovelyrouter.com/v1/tools/xiaohongshu-search/run \
  -H "Authorization: Bearer $LOVELY_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "input": {"operation": "search_notes", "keyword": "露营", "sort_type": "like_count_descending"},
    "max_results": 20
  }'

Each note is one item in items, in the tool's own format; run a small search first and look at the fields before you build on them. A run still going after 60 seconds returns 202; Run a tool shows how to poll.

Douyin works the same way. This run returns up to 50 of the most-liked coffee videos from the past week:

curl https://lovelyrouter.com/v1/tools/douyin-search/run \
  -H "Authorization: Bearer $LOVELY_API_KEY" \
  -H "content-type: application/json" \
  -d '{"input": {"keywords": ["咖啡"], "sort": "most_liked", "publishTime": "one_week"}, "max_results": 50}'

Xiaohongshu (RedNote). xiaohongshu-search returns notes for a keyword with operation set to search_notes. sort_type orders them by general, time_descending, or most likes, comments or saves (like_count_descending, comment_count_descending, collect_count_descending). Its input schema also lists other operations, among them get_note_detail and get_note_comments (with a note_id), get_user_info and list_user_notes (with a user_id), and search_products (with a keyword); choose one and fill only the fields it uses. Field descriptions are in Chinese and English.

Douyin. Douyin search returns videos for one or more keywords. Chinese, English, hashtags, brand names and emoji all work. Each keyword is searched on its own and duplicates across keywords are removed; a realistic keyword yields a few hundred unique videos. Sort by general, most_liked or latest, and limit publishTime to one_day, one_week or half_year.

Weibo. Weibo search runs one operation per call:

The numeric userId is in a profile URL (weibo.com/u/{ID}) and in the userId or user_id field of search results.

With the MCP server connected, one prompt does the first recipe: "Search Xiaohongshu for the 30 most-liked notes about 露营 and Weibo for posts about 露营装备, then summarize in English which brands and products people mention most. Set max_results on every run."

Each run bills per result: a note, a video or a post, plus a per-call fee on Weibo; the cards above show today's prices. Starting a run holds its maximum cost, per-call fee plus per-result fee times max_results, and settles to what it returned. A run that fails or times out pays only for the results it delivered, and a run refused at the start costs nothing.

There's no subscription. Top up from $5, and the same balance pays for the model that translates and summarizes. See pricing for every tool.