Stock News Monitor - Headlines, Buzz and Sentiment per Ticker

Give it a list of tickers and get one flat row per ticker: every headline Google News and Yahoo Finance published about the company in the last N days, deduplicated and counted, each one labelled positive / negative / neutral by a transparent finance lexicon, the share of Stocktwits posters tagging the symbol bullish, the last price with 1-day / 5-day / 30-day change and volume versus its 20-day average, and, on a Schedule, how many headlines are new since the previous run. Every source is public and key-free, read over plain HTTP: no browser, no login, no API keys. Pricing is pay-per-event, $5 per 1,000 tickers checked and $0.50 per 1,000 headline rows, no start fee, and nothing is charged for a ticker Yahoo does not know or for one where both news feeds failed.

Source on GitHub, MIT licensed. Not investment advice: this actor counts and labels public headlines, it does not predict prices, and nothing in its output is a recommendation to buy or sell anything.

Three things set it apart from the news scrapers in the Store:

• One row per ticker, priced per ticker. headline_count, headline_count_24h, the positive / negative / neutral split with an average score, the top sources, the five latest headlines and the Stocktwits and price context, in a single flat row you can drop into a sheet, a dashboard or an alert.
• Only headlines that are actually about the company. Yahoo's per-ticker feed pads with unrelated market stories (15 of the 17 NVDA items on 2026-10-29 did not mention Nvidia). A mention filter drops them and headlines_fetched tells you how many were dropped. Turn it off if you want everything.
• Run-to-run change detection. Turn on trackChanges, put the actor on a Schedule, and every run reports how many headlines are new since the last one and how the count moved. Free. Headline rows carry is_new_since_last_run.

What you get

One real ticker row for PLTR, from a live run on 2026-10-29 with the default settings (mode: "summary", 7-day window, Stocktwits 1 page, change tracking off):

{
  "row_type": "ticker",
  "ticker": "PLTR",
  "company_name": "Palantir Technologies Inc.",
  "exchange": "NASDAQ",
  "checked_at": "2026-10-29T19:14:11.188Z",
  "window_days": 7,
  "status": "ok",
  "headline_count": 66,
  "headlines_fetched": 72,
  "headline_count_24h": 19,
  "headlines_per_day": 9.4,
  "google_news_capped": false,
  "sentiment_label": "positive",
  "sentiment_score": 0.162,
  "sentiment_positive": 18,
  "sentiment_negative": 8,
  "sentiment_neutral": 40,
  "buzz_flag": null,
  "buzz_ratio": null,
  "stocktwits_bull_ratio": 0.625,
  "price_last": 193.61,
  "price_change_1d_pct": -1.63,
  "gdelt": null,
  "stocktwits": {
    "messages_sampled": 30,
    "bullish": 5,
    "bearish": 3,
    "unlabeled": 22,
    "bull_ratio": 0.625,
    "messages_per_hour": 14.2,
    "span_hours": 2.04,
    "watchlist_count": 325472,
    "newest_message_at": "2026-10-29T19:11:05.000Z",
    "oldest_message_at": "2026-10-29T17:08:36.000Z",
    "pages_read": 1
  },
  "price": {
    "currency": "USD",
    "exchange": "NasdaqGS",
    "last": 193.61,
    "prev_close": 196.81,
    "change_1d_pct": -1.63,
    "change_5d_pct": -1.32,
    "change_30d_pct": 3.51,
    "volume_last": 12892802,
    "volume_avg_20d": 22486000,
    "volume_ratio": 0.573,
    "as_of": "2026-10-29T19:11:45.000Z"
  },
  "top_sources": [
    { "source": "Yahoo Finance", "count": 9 },
    { "source": "TradingView", "count": 5 },
    { "source": "Benzinga", "count": 4 },
    { "source": "finance.yahoo.com", "count": 4 },
    { "source": "Investing.com Canada", "count": 3 }
  ],
  "top_headlines": [
    {
      "title": "Why J.P. Morgan Believes Palantir Stock Is Just Getting Going After a Lackluster 2026",
      "source": "finance.yahoo.com",
      "url": "https://finance.yahoo.com/m/ea73e0e6-1f97-3d96-9c86-0b7e89d1ec67/why-j.p.-morgan-believes.html",
      "published_at": "2026-10-29T18:00:00.000Z",
      "sentiment_label": "negative"
    },
    {
      "title": "Palantir Seen Beating Revenue Estimates, Raising Guidance, Oppenheimer Says",
      "source": "finance.yahoo.com",
      "url": "https://finance.yahoo.com/markets/stocks/articles/palantir-seen-beating-revenue-estimates-164852851.html",
      "published_at": "2026-10-29T16:48:52.000Z",
      "sentiment_label": "positive"
    },
    {
      "title": "Palantir Stock Drops: What's Going On?",
      "source": "Benzinga",
      "url": "https://news.google.com/rss/articles/CBMinAFBVV95cUxQVGpLYjRGQnpoT2JSNlV0YjR4WFg2WHhGMFZfU0huVzZNZEdyQURGLXJjWHRoMHMxUU5rMUFzRmhkQ2hoSEs1MGY5QzNkeU1QX0F5Qm13a0pnUEJPdFRiRHluekRsellSbE5nMFExYW5CMV95MHV1YVlTMkNzbW5DX2NmQWVrU0g4QkM1eWFTWkdqMGJ6aGpmeDBwSFc?oc=5",
      "published_at": "2026-10-29T16:12:16.000Z",
      "sentiment_label": "negative"
    }
  ],
  "change_tracking": null,
  "sources_ok": ["yahoo_search", "yahoo_rss", "stocktwits", "google_news", "yahoo_chart"],
  "warnings": [],
  "error": null
}

Reading it: in the last 7 days the two feeds returned 72 headlines tagged PLTR, 66 of which name Palantir; 19 of those were published in the last 24 hours, about 9 a day over the week. 18 read positive and 8 negative by the lexicon, so the week nets mildly positive. Of the 30 most recent Stocktwits posts, 5 were tagged bullish and 3 bearish; the rest carried no tag. The stock was down 1.6% on the day on roughly half its usual volume.

One real headline row (mode: "headlines" or "both"), the first for the same ticker:

{
  "row_type": "headline",
  "ticker": "PLTR",
  "headline_key": "960376a717bd257e",
  "title": "Palantir Seen Beating Revenue Estimates, Raising Guidance, Oppenheimer Says",
  "source": "finance.yahoo.com",
  "url": "https://finance.yahoo.com/markets/stocks/articles/palantir-seen-beating-revenue-estimates-164852851.html",
  "published_at": "2026-10-29T16:48:52.000Z",
  "origin": "yahoo_finance",
  "mentions_company": true,
  "sentiment_label": "positive",
  "sentiment_score": 1,
  "positive_terms": ["beating", "raising guidance"],
  "negative_terms": [],
  "is_new_since_last_run": null,
  "scraped_at": "2026-10-29T19:14:11.188Z"
}

positive_terms and negative_terms are the exact lexicon hits that produced the score, so you can audit any label, filter on specific terms, or re-score the title with your own model.

Use cases

Daily market recap, newsletter or video research. Run mode: "both" on your coverage list every morning. The ticker rows tell you which names had a loud day and which way the headlines leaned; the headline rows give you the deduplicated reading list with links. 20 tickers a day is 20 checks plus about 1,500 headline rows, roughly $0.85 a day.

Investor-relations and PR monitoring. Put your own ticker and your peers on an hourly Schedule with trackChanges: true and mode: "both". change_tracking.new_headlines on the ticker row and is_new_since_last_run on the headline rows give you only what is new since the last hour, ready for a Slack or email integration. Media-monitoring suites charge hundreds a month for this.

A research journal for your own watchlist. Run it daily on the names you follow and keep the dataset. In a few months you have, for every day, what was written, how it leaned and what the price did next, which is the raw material for an honest look at whether the news you react to actually meant anything. It will not tell you what to buy. It will show you what you were reading when you decided.

Input

Field	Type	Default	Notes
tickers	string[]	["NVDA","AAPL","TSLA"] (prefill)	Yahoo Finance symbols: NVDA, BRK-B, ASML.AS, 7203.T, ^GSPC. A leading $ is stripped, case is ignored, duplicates are removed. Soft cap 1,000 per run.
mode	summary / headlines / both	summary	One row per ticker, one row per headline, or both in one dataset (row_type tells them apart).
windowDays	1-30	7	How far back headlines are collected and counted.
requireMention	boolean	true	Keep a headline only when its title contains the ticker or a distinctive word of the company name. headlines_fetched always reports the unfiltered count.
maxHeadlinesPerTicker	1-500	100	Cap on headline rows per ticker in headlines / both mode, newest first. The ticker row always counts every headline collected.
stocktwitsPages	0-10	3	Pages of 30 recent Stocktwits posts to aggregate. 0 skips Stocktwits. Only counts are emitted, never posts or users.
includePrice	boolean	true	Last price, 1d / 5d / 30d change and volume vs 20-day average from Yahoo Finance.
includeGdelt	boolean	false	Experimental: 30-day daily mention-volume series from the GDELT DOC API for buzz_ratio / buzz_flag. GDELT throttles hard (see limits), so it is off by default.
trackChanges	boolean	false	Remember the headlines seen per ticker and report what is new next run. Free.
stateStoreName	string	stock-news-monitor-state	Key-value store that holds the per-ticker snapshot. Use one name per watchlist.
proxyConfiguration	proxy	Apify datacenter	All sources are public, but Yahoo and Stocktwits throttle shared datacenter IPs. Switch to RESIDENTIAL if a run reports many 403 / 429 warnings.
requestDelayMs	100-10000	500	Minimum gap between two requests to the same host across the whole run. GDELT is always paced at 10 s.
maxConcurrency	1-8	4	Tickers processed in parallel.

Output

Ticker row (row_type: "ticker", event ticker-check)

Field	Meaning
ticker, company_name, exchange	The symbol as normalised, and what Yahoo Finance calls it.
checked_at, window_days	When the row was built and the window it covers.
status	ok, not_found (Yahoo knows no such symbol, not charged) or error (both news feeds failed, not charged).
headline_count	Deduplicated headlines inside the window after the mention filter.
headlines_fetched	The same before the mention filter.
headline_count_24h, headlines_per_day	Headlines published in the last 24 hours, and the average per day over the window.
google_news_capped	true when Google News returned its maximum of 100 items, so the window is only sampled (see limits).
sentiment_label, sentiment_score	Average headline score over the window and the label it implies (positive above 0.15, negative below -0.15, else neutral). null with no headlines.
sentiment_positive, sentiment_negative, sentiment_neutral	How many headlines fell in each bucket.
stocktwits_bull_ratio, stocktwits	bullish / (bullish + bearish) over the sampled posts, plus the counts, posts per hour, the span sampled and the symbol's watchlist count. null when the symbol is not on Stocktwits (most non-US listings).
price_last, price_change_1d_pct, price	Last regular-market price, the change against the previous close, and the full context object (5d / 30d change, volume and its 20-day average, currency, exchange, timestamp).
buzz_flag, buzz_ratio, gdelt	Only with includeGdelt: latest GDELT day's mention volume divided by the mean of the previous days, flagged unusual_high (2x or more), elevated (1.3x), normal or quiet (0.5x or less). null otherwise.
top_sources	The five publishers with the most headlines in the window.
top_headlines	The five most recent headlines with source, link, time and label.
change_tracking	With trackChanges: previous_run_at, previous_headline_count, new_headlines (never seen in an earlier run) and headline_count_change_pct. null otherwise.
sources_ok, warnings	Which of the six sources answered, and one line per source that did not (the ticker is still delivered).
error	Only on not_found / error rows.

Headline row (row_type: "headline", event headline-row)

Field	Meaning
ticker, headline_key	The ticker and a stable 16-hex id derived from the normalised title (the same story from both feeds gets one key).
title, source, url, published_at	The headline, the publisher, the link and the publication time. Yahoo links are direct; Google News links are Google redirect URLs.
origin	google_news or yahoo_finance: which feed delivered it first.
mentions_company	Whether the title names the company or ticker. Always true when requireMention is on.
sentiment_label, sentiment_score	positive / negative / neutral and (positive - negative) / (positive + negative) over the matched terms.
positive_terms, negative_terms	The lexicon hits, with not ... prefixes where a negator flipped a term.
is_new_since_last_run	With trackChanges: true when no earlier run saw this key. null otherwise.
scraped_at	When the row was built.

Run metrics (requests, blocked requests, failures per source, tickers by outcome, charge counters) are stored in the OUTPUT record of the run's default key-value store.

Pricing

Event	Price	Charged when
ticker-check	$0.005	A ticker row with status: "ok" is pushed (summary and both modes). Never for not_found or error rows.
headline-row	$0.0005	A headline row is pushed (headlines and both modes), deduplicated by title within the run.

No start fee. 1,000 tickers in summary mode cost $5.00. With the defaults a ticker is 5 to 7 requests (search, Google News, Yahoo RSS, chart, 3 Stocktwits pages) and about 250 KB of transfer, which is under $0.001 of platform usage on datacenter proxies, so the check price is almost entirely the product.

The maximum total charge you set on the run is respected: the actor stops taking new tickers once the budget would be exceeded and reports how many were not delivered. A ticker is never charged without its row in the dataset.

Limits, honestly

• Google News returns at most 100 items per query. For heavily covered names (NVDA, AAPL, TSLA) a 7-day window hits that cap, google_news_capped is true, and headline_count is a sample, not a census. The sample also rotates slightly between runs, which inflates new_headlines for capped tickers until a few runs have been seen. Use a shorter windowDays for busy names if you need completeness.
• Sentiment is a keyword count, not understanding. It uses a finance lexicon of about 300 words and phrases with simple negation. It sorts a day's headlines into "mostly bad" / "mostly good" well enough and gets individual headlines wrong regularly ("gained access" counts as a gain). Every row carries the matched terms so you can see exactly why. In the live runs about a third of headlines received a label and the rest stayed neutral.
• Stocktwits sentiment is self-reported by the fraction of posters who tag a post, typically a third. bull_ratio is that fraction's lean, nothing more, and the symbol stream only exists for US-listed names.
• Price context is context. The last price is Yahoo's regular-market quote at run time, which during the session is intraday. Volume ratios during the session compare a partial day with full days.
• GDELT (includeGdelt) is experimental. The free DOC API allows about one request every 5 to 10 seconds per IP and imposes a cooldown after a burst; from a shared IP it often refuses. When it does, the buzz fields are null, the warning says so and the ticker is delivered and charged normally. It also lags one to two days.
• Yahoo Finance and Stocktwits throttle datacenter IPs. Blocks are retried on a fresh proxy session; a source that stays blocked is reported in warnings and the ticker is still delivered from the sources that answered. If every news source is blocked for every ticker, the run fails with a clear message instead of finishing green with an empty dataset.
• Google News links are redirects (news.google.com/rss/articles/...). The publisher's home is in the source; resolving the final article URL would cost one request per headline and is not done.
• Not investment advice. The author is not a licensed advisor. The actor reports what was published; it has no opinion about what a security is worth.

Change tracking

With trackChanges: true the actor keeps one record per ticker in the key-value store named by stateStoreName: the time of the last run, its headline count and the keys of every headline seen with the date it was first seen (pruned after 90 days). On each run, a headline whose key is absent from the snapshot is new. The snapshot is written only after the ticker's rows were delivered, so a run that stops at the charge limit re-diffs the undelivered tickers next time. Use a different store name for each watchlist; the first run on a new store reports every headline as new.

Changelog and support

Source-side observations (endpoints, payload shapes, throttling) are recorded in the changelog with the date they were verified. Issues are answered within 24 hours.