China Financial Flash News Monitor — CLS 财联社 电报

A keyless real-time feed of China's most-watched market newswire — the CLS (财联社) "电报" telegraph flash stream. Every item comes with its associated A-share tickers and a bull/bear sentiment score, so you can wire market-moving headlines straight into an event-driven model. Built for quant / systematic funds, event-driven desks, and China research.

No login. No API key. No CAPTCHA.

Why CLS

CLS is the fastest, most-cited China financial newswire — the flash that moves A-shares intraday. This Actor gives you that stream as clean structured records, each linked to the stocks it names, without a terminal.

What you get

Field
id, publishedAt, ctime	the flash + its timestamp
title, content	headline + body
associatedStocks	the A-share tickers the flash names (code, name)
sentiment, sentimentScore	bull/bear label + score in [-1, 1]
readingNum, level, url	engagement, importance, direct link

A run rollup (item count, sentiment split, items-with-stocks) is written to the key-value store SUMMARY.

Built for a schedule

Turn on deltaMode + an Apify Schedule. Flash news never stops, so schedule this from every few minutes to hourly — each run returns only items new since the last (deduped on id), a gap-free feed that charges only for new flashes.

Set stockCodes to keep only the flashes that name your watchlist.

Example input

{
    "category": "",
    "stockCodes": ["600519", "000858", "300750"],
    "sentimentAnalysis": true,
    "deltaMode": true,
    "deltaStateKey": "my-watchlist"
}

Leave category empty for the full 全部 stream, or use red for market-moving items only. Leave stockCodes empty for the whole newswire.

Pricing

Pay-per-event: $0.008 per flash item. In deltaMode you're charged only for genuinely new items since your last run.

Notes & honest limitations

• Many flashes are macro / sector / policy and name no individual stock — a stockCodes filter will (correctly) drop those.
• Sentiment is a transparent bull/bear lexicon score labelled heuristic-lexicon, not a trained model — run your own model on the raw content for production signals.
• Aggregated public news metadata — no personal data.
• Found a field you need or a bug? Open an issue on the Actor page — I typically respond within 48 hours.