How do I let my AI read Zhihu, Bilibili and Weibo? (2026)
1 sentence with your key connects all three: paste it into Claude Code, Cursor or Codex, the agent sets up DataXAPI's MCP server, and then it searches Zhihu, Bilibili and Weibo when you ask, at $0.0020 a call, with no account or cookie of yours. DataXAPI reads Zhihu with 16 endpoints, from $0.0020 a call; a failed call is not charged.
Updated 2026-11-04
Connect by MCP
Paste this into the agent: “Read https://dataxapi.com/agents.md and set up DataXAPI in this agent. My API key is dx_…; use it for every DataXAPI call.”
Or from a terminal: npx dataxapi setup writes the entry for these clients and installs the base skill. It does not write your key: put that in the environment variable DATAXAPI_API_KEY.
Or add a Skill
Each platform has a Skill that lists its endpoints, its prices and how to call them: “Install the DataXAPI skill dataxapi-zhihu from https://dataxapi.com/skills/dataxapi-zhihu.zip.” For the others, use dataxapi-bilibili and dataxapi-weibo in both places. The Skill itself is free, you pay per call, and it needs the connection above and a key.
What to ask, and what it comes to
Finding an endpoint and reading its parameters are free; only call is charged, and the agent can see the price first. Give it its own key with a limit.
| Ask | Calls | Comes to |
|---|---|---|
| “Search Zhihu for 咖啡机推荐, read the answers to 3 of the questions, and tell me where they agree.” | 4 × $0.0020 | $0.008 |
| “Find Bilibili's most played videos on 机械键盘, and tell me what the comments under the first 5 complain about.” | 6 × $0.0020 | $0.012 |
| “Search Weibo for this brand, 3 pages, read 5 long posts in full, and group what people say.” | 8 × $0.0020 | $0.016 |
Zhihu's own MCP
Zhihu's data platform has its own MCP. Its document, read on 2026-11-04, says that in Codex, WorkBuddy, Claude Code, Cursor and others it searches Zhihu and the wider web, reads the hot list and uses Zhihu's own answer agent, with an Access Secret from your account there. DataXAPI gives whole answers, comments and authors, and Bilibili and Weibo on the same key.
What it can and cannot read
Zhihu: zhihu.posts.search for answers and articles, zhihu.questions.answers.list for the first 20 answers to a question, zhihu.answers.comments.list for comments. Answers run long, and view=compact cuts each text to 500 characters, which helps when the AI is only sifting.
Bilibili: bilibili.videos.search gives title, uploader, plays, likes, saves and comment counts, and bilibili.videos.comments.list the comments. It cannot read what is said in a video, and there is no endpoint for danmaku text: view=full gives only their count, in ext.danmaku.
Weibo: weibo.posts.search gives posts newest first, with long ones cut short; weibo.posts.get has the whole text. On all three it reads public data only and cannot post, like or comment.
Endpoints
Each with its price; every page has the parameters, the fields and a recorded response.
Sources
What this page says of other services is from their own pages and documents, read on 2026-11-04. Prices and features change: check the source before you decide.