36Kr China Tech & Startup News Scraper

Scrapes 36Kr (36氪) — China's equivalent of TechCrunch — and extracts article metadata, funding-round announcements, and author data. Covers all major sections: 创投 (venture/startup), AI, 科技 (tech), 出海 (overseas expansion), 新经济 (new economy), and more.

36Kr is where Chinese funding rounds often break before Western press picks them up. If your workflow depends on early signals from China's startup ecosystem, this is the source.

What You Get

• Article ID, URL, title, and summary for each piece
• Author name, publication timestamp (ISO-8601), and section label
• Tags, hero image, read count, and comment count when available
• Funding round label (Series A, B, etc.), amounts in CNY and USD — parsed from article leads when 36Kr surfaces them
• Optional full body extraction (HTML + plain text) when you need the full text, not just the summary

Features

• Section filtering — target specific verticals (创投, AI, 科技, 出海) instead of crawling everything
• Date range filtering — skip articles before a given date
• Full body mode — fetches each article page to extract content and enriches funding data from the full text
• Residential proxy routing — routes through DataImpulse residential proxies to avoid ByteDance's bot detection on listing pages

How It Works

1. Seeds the requested section listing pages (/information/web_<slug>/)
2. Parses window.initialState embedded JSON — 30 articles per page
3. Normalises each item from 36Kr's internal data shape into the output schema
4. Paginates using hasNextPage signal until maxItems is reached
5. Optionally fetches each article page for full body HTML and investor details

Input

Field	Type	Default	Description
maxItems	integer	10	Maximum articles to scrape
sections	array	["web_news"]	Section slugs to crawl. Available: web_news, web_zhichuang, web_innovate, web_ai, web_tech, web_chuhai
startDate	string	—	ISO-8601 date. Skip articles published before this date
extractBody	boolean	false	Fetch each article page for full body HTML/text

Input Example

{
  "maxItems": 50,
  "sections": ["web_zhichuang", "web_ai"],
  "startDate": "2025-01-01",
  "extractBody": false
}

Output

Each result is one article record.

Field	Type	Description
article_id	string	19-digit article ID from the URL
url	string	Canonical article URL
title	string	Article title (Chinese)
summary	string	Lead paragraph
body_html	string	Full article HTML (only when extractBody: true)
body_markdown	string	Full article as plain text (only when extractBody: true)
author	string	Author display name
publish_time	string	ISO-8601 publication timestamp
section	string	Section label (e.g., 创投, AI, 科技)
tags	string	JSON array of subject tags
companies_mentioned	string	JSON array of company entities (from widget data when extractBody: true)
funding_round	string	Round label when article covers a funding event (e.g., "Series B")
funding_amount_cny	number	Amount in CNY when surfaced in lead
funding_amount_usd	number	Amount in USD when surfaced in lead
lead_investors	string	JSON array of lead investor names (when extractBody: true)
hero_image	string	Hero image URL
read_count	integer	View count when available
comment_count	integer	Comment count when available

Sample Output

{
  "article_id": "2492318105786505",
  "url": "https://36kr.com/p/2492318105786505",
  "title": "某AI公司完成10亿元A轮融资",
  "summary": "该公司专注于工业视觉检测，本轮融资由红杉领投。",
  "body_html": null,
  "body_markdown": null,
  "author": "36氪创投",
  "publish_time": "2025-05-30T03:14:00.000Z",
  "section": "创投",
  "tags": "[\"人工智能\",\"融资\",\"工业互联网\"]",
  "companies_mentioned": "[]",
  "funding_round": "Series A",
  "funding_amount_cny": 1000000000,
  "funding_amount_usd": null,
  "lead_investors": "[]",
  "hero_image": "https://img.36krcdn.com/hsossms/20250530/v2_image.jpg",
  "read_count": 4200,
  "comment_count": 12
}

🔍 FAQ

Is this only for the news section? No. Use the sections input to target specific verticals. 创投 (web_zhichuang) covers startup and funding news specifically, which is often the highest-signal content for VC and deal-flow use cases.

How fresh is the data? The listing pages update continuously. A run against web_news with maxItems: 30 captures the last hour or two of posts.

Does it extract full article text? Set extractBody: true. The actor fetches each article page separately and returns body_html and body_markdown. Expect longer run times proportional to maxItems.

How are funding amounts parsed? From Chinese-language patterns in article titles and summaries (e.g., "完成X亿元A轮融资"). When extractBody: true, the full article body is also scanned for more precise figures. This is heuristic — verify amounts for critical applications.

What sections are available? web_news (all), web_zhichuang (创投/VC), web_innovate (新经济), web_ai (AI), web_tech (科技), web_chuhai (出海). The site has additional sections not listed here — their slugs follow the same web_<name> pattern.

What about the gateway API at gateway.36kr.com? It requires signed tokens and is not used. The SSR listing pages embed all article data needed to populate the schema.