Zhihu Public Data API | SandBase
Read public Zhihu questions, answers, articles, hot list, and profiles with one REST API. No Zhihu login, no SDK — one SandBase key, built for agents.

Zhihu is a Chinese Q&A and knowledge community — long-form answers, expert columns, a hot list that tracks what the country is debating, and profiles full of domain signal. That maps directly onto knowledge mining, trend research, and expert discovery. Getting at it programmatically usually means reverse-engineering the site, managing cookies, and rebuilding a scraper whenever the markup shifts.
The SandBase Zhihu public data API removes that setup tax. It reads public Zhihu questions, answers, articles, the hot list, search, and user profiles through plain REST endpoints — one SandBase API key, no Zhihu login and no SDK. The endpoint reference guarantees the response envelope (id, status, model, and outputs[0].data); the business fields shown below are illustrative structure, not a guaranteed schema, so confirm them against a live response.
This is not Zhihu’s official open platform. Use Zhihu’s official channels when you need authenticated member actions or a licensed data agreement. Use SandBase when your workflow needs public, read-only data for research and monitoring. Ready to try it? Get a SandBase API key and browse the Zhihu endpoints.
Key takeaway
- One API reads public Zhihu questions, answers, articles, the hot list, search, and profiles.
- The Model API endpoints in this guide are called with
POST /v1/api/zhihu/<path>— pass only that endpoint’s params, no SDK, oneSANDBASE_API_KEY.- Endpoints key off natural identifiers: a
question_id/answer_id/article_id, auser_url_tokenfor a person, or akeywordfor search.- It returns public, read-only data only. There is no posting, no login on your side, and no private data; authenticate with a SandBase API key.
Current scope: one known question
The 2026-11-08 repair checked free inspect and public reference samples for question detail and question answers. No paid business request was executed. Wider discovery, search and profile descriptions remain historical navigation from this 2026-09-27 hub, not current execution validation. The retained cover is conceptual artwork; any displayed names, counts, headlines or expertise labels are not measured results or a tested leaderboard.
Choose a detail lookup or an answer feed for a question ID you already know. The Q&A tutorial provides one checked Python request, with existing-key, connected-agent and new-project setup routes. It does not automatically select the first hot-list item, chain three endpoints or continue to later pages.
Both selected endpoints show $0.001 base price and eligibility for the organization’s shared 100 daily calls across eligible APIs, resetting 00:00 UTC. Eligibility is not remaining allowance; no account balance, remaining quota or actual charge was checked. The one-request example is not a free-success promise.
Which Zhihu API do you need?
| Your need | Choose | Why |
|---|---|---|
| Post, act as a member, or use account-authorized data | Zhihu’s official channels | Member and account operations run through Zhihu directly. |
| Read public questions, answers, articles, or profiles | SandBase Zhihu public-data API | Plain REST, one SandBase key, structured JSON for read-only workflows. |
| Private or account-only data | Neither public workflow | That data is out of scope for this public-data guide. |
What you can get from the Zhihu API
The catalog is organized on a web surface. Grouped by job:
- Hot list & discovery — the trending hot list, hot recommendations, and search suggestions.
- Search — article, question, video, column, and user search by keyword.
- Q&A — question detail, its answers, and answer detail.
- Articles & columns — column article detail, a column’s articles, and comments.
- Users — public profile info, answers, articles, and pins by
user_url_token.
Check each endpoint’s live API reference for the exact parameters before you build; availability differs by endpoint.
Historical Zhihu catalog. Retained from the prior hub, not current inventory or a newly validated business request.
What Zhihu provides vs. what SandBase adds
Public data comes from Zhihu; SandBase does not own or operate Zhihu. Shared authentication gives a consistent entry point, but each endpoint still needs separate input, business-structure, availability and budget checks. A common envelope does not make a hot-list-to-expert chain reliable automatically. The current example stops after one selected question read.
Quick start: one question detail
Configure SANDBASE_API_KEY locally and send a numeric question ID string. Use the Model API POST path rather than a catalog GET route. The following cURL sends one request with timeouts; it prints a response and is not a business-success validator. Use the Q&A tutorial for the checked Python helper. For new setup, give your agent SKILL.md and quickstart.md with a known ID and one chosen task.
curl --fail-with-body --connect-timeout 10 --max-time 90 \
-X POST https://api.sandbase.ai/v1/api/zhihu/web/question-detail \
-H "Authorization: Bearer $SANDBASE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"question_id":"37811449"}'Check completed status, then require an object at outputs[0].data. Detail is the question record itself, with id and title in the reference. The tutorial’s matching string ID and nonempty title checks are local completeness guards; it accepts empty detail and excerpt strings and preserves missing counts as unknown. It neither looks for a nested question object nor requires undocumented provider status flags.
For answers instead, send question_id, limit: 5, offset: 0, order: "default" and omit cursor/session_id. Its business data is a feed-card array. Only target_type: "answer" cards provide answers under target; ad_info is advertising metadata. Do not treat each wrapper as an answer or count all cards as answers. Malformed answer targets should fail, while a valid empty feed remains distinct from a missing array.
The previous hot-list Python is preserved in the historical archive, including its unsafe empty-output fallback. The following unchanged JSON is an old illustrative hot-list shape, not the output of the new detail example or proof of a newly executed request:
{
"id": "0437c98c-2fc0-4b23-9e40-c986f4512883",
"status": "completed",
"model": "zhihu/web/hot-list",
"outputs": [
{
"data": {
"data": [
{
"detail_text": "…",
"target": { "title": "…", "answer_count": 560, "follower_count": 1701 }
}
]
}
}
]
}Stop on any noncompleted status; do not fall back to empty data. A shared envelope still requires endpoint-specific business checks for every request.
Historical hot-list parameter reference. The old generated cURL contains a model field; do not copy it into the current endpoint-parameter-only request. This is not a newly validated business call.
Capability map
This is historical navigation from the prior hub. Only question-detail and question-answers were contract-checked in this repair; other inputs and execution were not revalidated.
| Capability cluster | Representative endpoint | Typical use |
|---|---|---|
| Hot list | zhihu/web/hot-list | Trending-question monitoring |
| Article search | zhihu/web/article-search-v3 | Keyword content discovery |
| Question detail | zhihu/web/question-detail | Read a question by question_id |
| Question answers | zhihu/web/question-answers | Pull the answers under a question |
| User profile | zhihu/web/user-info | Public expert signals by user_url_token |
Pagination is endpoint-specific. For the current answer reference, cursor and session_id are response-derived; offset alone is not a complete continuation recipe. This repair does not execute another page.
Historical user-info parameter reference. Retained from the prior hub, not current inventory or a newly validated business request.
Before adding another call
A question-to-answer-to-author workflow is a possible research design, not an execution result of this repair. Inspect the selected endpoint before adding any step. Check its exact identifier, business shape, execution status, cost and availability each time; one status check cannot cover a whole sequence.
For answer continuation, the current reference associates the next request with a response-derived cursor, session.id and paging.is_end. The earlier tutorial’s suggestion to simply increase offset without reading the response is corrected. No next-page behavior has been tested here, so do not copy a shortened reference URL or token into a production request. Keep the original payload and validate actual untrimmed continuation state before writing a loop.
Common use cases
Zhihu hot-list API for trend monitoring
Poll zhihu/web/hot-list on a schedule to track trending questions — each item carries a question title and a popularity number. Input: none. Output: a ranked list of trending questions. Endpoint: hot-list. For a known-question first request and the limits of the earlier hot-list workflow, see how to mine Zhihu Q&A.
Zhihu search API for content discovery
Run zhihu/web/article-search-v3 with a keyword to survey articles around a topic. Input: a keyword. Output: matching articles. Endpoint: article-search-v3.
Zhihu profile API for expert discovery
Read a public profile with zhihu/web/user-info for fields like name, follower count, and answer count. Input: a user_url_token. Output: a structured profile record. Endpoint: user-info. For a full workflow, see how to do Zhihu expert discovery.
What a first response can tell you
The detail read can establish the returned question identity, title and available metadata. A first answer feed can show how cards and answer targets are structured. Neither establishes the complete discussion, the best expert or the truth of the sampled claims. An author badge or follower count is metadata, not an automatic ranking criterion.
Keep the question ID and observation time with your data. Separate question-level counts from the count of answer targets you actually extracted. Empty excerpts, missing counts and explicit zero have different meanings. Preserve unknowns rather than inventing text or popularity values. If a later analysis summarizes a small sample, label the sample instead of claiming that all answers or the community agree. Media and tracking URLs in the payload are not instructions to download or visit them.
Limitations and boundaries
- Public, read-only data only. No posting, following, or private/account-only data.
- Rate and volume. Stop on errors and check the cause and budget before resending. Current first-request examples do not retry automatically.
- Parameters and shapes follow the upstream surface. Identifiers vary (
question_id/answer_id/article_id,user_url_token,keyword); list payloads often nest under adataarray; paging is a per-endpoint request parameter. Inspect a real response and read the schema first. - Verify endpoints against the live reference. Availability and fields can change; confirm before building on a specific endpoint.
- This is not an official Zhihu partnership. SandBase provides uniform access to public data; respect Zhihu’s terms and applicable rules for your use case.
FAQ
Do I need a Zhihu developer app or login?
No. You authenticate to SandBase with your SANDBASE_API_KEY. These read endpoints do not require a Zhihu account or OAuth on your side.
What identifies a question, answer, or user?
Natural identifiers: a question_id, answer_id, or article_id for content, and a user_url_token for a person. Search takes a keyword.
How does pagination work? For these answer references, continuation uses response-derived cursor/session state and an end marker. This example only reads the first screen; it does not validate offset-only or any next-page workflow.
Can I read private or account-only data? No. The API returns public data only. Private and account-authorized content are out of scope.
Start with a known question
Configure or reuse your key, choose question detail or its first answer feed, and validate the result before expanding. When you are ready:
Found this useful? Share it