Apify Financial News

Financial News Intelligence

Discover and extract financial news for portfolio companies via Apify Actors. Two modes:

• Single company — news scan for one company.
• Portfolio scan — same pipeline run across multiple companies.

33 Tier 1 sources organized in 4 categories (Global / Pan-European / Institutional / CEE Local). Tier 2 = broad Google News fallback for unverified domains.

Estimated cost: $0.10–0.50 per single company, $1–5 per full portfolio scan.

Prerequisites

• Apify access — preferred: apify CLI (npm install -g apify-cli && apify login); fallback: Apify MCP connector (call-actor tool). CLI is faster and preferred when both are available.
• Python 3 + pip install readability-lxml lxml (for the extract_and_clean.py post-processor)
• Companies data at ${CLAUDE_PLUGIN_ROOT}/data/companies.json

${CLAUDE_PLUGIN_ROOT} is the plugin's root directory (where .claude-plugin/ lives). It is resolved automatically by Claude Code when the plugin is installed, or set to the --plugin-dir path during development.

Workflow checklist

Copy this and tick boxes as you progress:

Task Progress:
- [ ] Step 0: Verify prerequisites — try `apify --version && apify info`; if unavailable, check for `call-actor` MCP tool; if neither, tell user to install apify CLI or Apify MCP connector. Also verify: `python3 -c "from readability import Document; print('OK')"` (install: `pip install readability-lxml lxml`)
- [ ] Step 1: Build queries (look up company in data/companies.json or construct manually)
- [ ] Step 2: Discovery — pick 8-12 sources by region, run 2-phase Google News
- [ ] Step 3: Dedup + route (Tier 1 = whitelisted domain → verified extractor; Tier 2 = broad → rag-web-browser)
- [ ] Step 4: Extract & clean (run extractor, then extract_and_clean.py)
- [ ] Step 5: Output Tier 1 + Tier 2 tables to user

Constraints

Allowed Apify Actors (exhaustive — do NOT use others)

Do NOT use any other actor. Do NOT use WebSearch, WebFetch, or browser tools.

Extractor routing (mandatory — do NOT substitute)

Full per-source config: reference/SOURCE_CONFIGS.md. Machine-readable: data/sources.json.

Pipeline

Step 1: Build queries

Look up company in ${CLAUDE_PLUGIN_ROOT}/data/companies.json. Key fields under queries: gnews_en, gnews_cz, bloomberg.

For non-portfolio companies, construct manually: quoted full legal name + ticker OR variant + geographic qualifier.

Query rules:

• Use "InPost SA" not InPost (quoted full names avoid false positives — see reference/EUROPEAN_COMPANIES_GUIDE.md)
• Per-source site: operator: site:bloomberg.com "InPost SA" OR "INPST"
• FT tip: use site:ft.com/content/ (bare ft.com returns stock-data pages)
• Valid timeframes: "1h", "1d", "7d", "1y", "all" (NOT "30d")
• Tickers < 4 chars: always pair with full company name
• ALWAYS set decodeUrls: true on Google News input

Step 2: Discovery

Do NOT search all 33 sources. Pick 8–12 based on company region.

Regional priorities

Two-phase Google News strategy

Phase 1 — Targeted (per priority source, with site:):

apify call data_xplorer/google-news-scraper-fast \
  --input '{"keywords":["site:bloomberg.com \"InPost SA\" OR \"INPST\""],"maxArticles":10,"timeframe":"7d","region_language":"US:en","decodeUrls":true,"proxyConfiguration":{"useApifyProxy":true,"apifyProxyGroups":["RESIDENTIAL"]}}' \
  --user-agent apify-awesome-skills/apify-financial-news \
  --output-dataset > discovery_bloomberg.json

Phase 2 — Broad (always run, no site: operator). Classify results by domain in Step 3 — whitelisted → Tier 1, others → Tier 2.

CEE local-language discovery: For CEE companies, run additional queries with region_language set to CZ:cs, PL:pl, HU:hu, BG:bg, SK:sk. See region_language field per source in data/sources.json.

EC Press Corner — direct REST API (no Actor)

curl -s "https://ec.europa.eu/commission/presscorner/api/documents?reference=IP/26/614&language=en"

Parse IP_XX_NNN reference IDs from Google News titles to construct API calls.

RSS / Headline discovery (optional)

For sources with RSS, you can supplement GNews with louvre/rss-news-aggregator (max 10 feeds per run; split into batches). For 6 sources, rodrigo_pacelli/headline-news-scraper works (CNBC, SCMP, Nikkei, Caixin, Zawya tag-pages-only, Handelsblatt). RSS/headline output is unfiltered — filter client-side by company name in title/description. Full feed list: reference/PIPELINE_DETAIL.md.

Step 3: Dedup & route

1. Collect URLs from all discovery runs. 2. Classify by domain: whitelist (33 Tier 1 sources) → Tier 1; other → Tier 2. 3. Filter non-article URLs: /quote/, /stock/, /sitemap, /author/, /tag/, /key-metrics/, /newsletters/, /topic/, /profile/, redirectUrl=. 4. Filter by company name/ticker in title. 5. Deduplicate (URL normalize). 6. Route Tier 1 URLs to verified extractor per routing table. 7. Route Tier 2 URLs to apify/rag-web-browser.

Low-coverage fallback (< 3 articles): broaden timeframe to "1y", run Phase 2 if skipped, try local-language queries.

Step 4: Extract & clean

Run the extractor per the routing table. For rag-web-browser calls, use outputFormats: ["html"].

After extraction, run extract_and_clean.py on the dataset to strip nav/menus/footers via readability-lxml. The script auto-detects format: HTML cleaned via readability-lxml, already-clean output (Bloomberg scraper, workhard3000) passes through.

DATASET_ID=$(apify call apify/rag-web-browser \
  --input '{"query":"<ARTICLE_URL>","maxResults":1,"outputFormats":["html"],"requestTimeoutSecs":40,"proxyConfiguration":{"useApifyProxy":true,"apifyProxyGroups":["RESIDENTIAL"]},"removeCookieWarnings":true}' \
  --user-agent apify-awesome-skills/apify-financial-news \
  --json | jq -r '.defaultDatasetId')

python3 ${CLAUDE_PLUGIN_ROOT}/skills/apify-financial-news/reference/scripts/extract_and_clean.py "$DATASET_ID"

The --json flag returns run metadata including defaultDatasetId. Errors produce non-zero exit code so the pipeline fails fast.

Step 5: Output

Two tables to the user — Tier 1 (verified) and Tier 2 (broad). Per article: source, title, author, date, char count, URL.

## News Intelligence: InPost (INPST.AS) — Last 7 days

### Tier 1 — Verified Sources (2)
| Source | Title | Author | Date | Chars | URL |
|--------|-------|--------|------|-------|-----|
| bloomberg.com | InPost Readies AI Shopping Assistant | K. Krasuski | 2026-03-19 | 7,577 | … |

### Tier 2 — Broad Discovery (5)
| Source | Title | Date | Chars | URL |
|--------|-------|------|-------|-----|
| seekingalpha.com | InPost expands parcel locker network | 2026-03-18 | 1,200 | … |

*Sources: 6 verified queried | 5 broad | Cost: $0.15*

Macro context

For country-level economic context, use these alongside news scans:

• ING Think (think.ing.com) — daily CEE FX/rates via apify/rag-web-browser (~11K chars). Best free open-access CEE macro source.
• IMF (imf.org) — Article IV concluding statements via apify/rag-web-browser (~22K chars).
• ECB (ecb.europa.eu) — already in Tier 1.
• Central banks (ČNB, NBP, MNB, BNB, NBS) — direct URL extraction; see reference/MACRO_SOURCES.md.

Critical gotchas

• ALWAYS `decodeUrls: true` in Google News (encoded redirects break ALL extractors).
• "InPost" is ambiguous — matches "post-Maduro". Use "InPost SA" for precision.
• RSS max 10 feeds per run — split into batches.
• Zawya headline-scraper returns tag pages, NOT articles — exclude from headline runs.
• Reuters needs RESIDENTIAL proxy — without it, returns 386 chars.
• Reuters: rag-web-browser returns 0 chars on ~60% of URLs — use workhard3000 only.
• Morningstar `.co.uk` URLs fail with workhard3000 — use .com URLs for extraction.
• WSJ livecoverage pages fail — skip URLs matching wsj.com/livecoverage/.
• FT: rag-web-browser returns 16 chars ('Client Challenge') — use workhard3000.
• Forbes: workhard3000 returns 0 chars — use rag-web-browser.
• Barron's / MarketWatch cost $0.25/page — use selectively for high-value articles.
• Caixin URLs must be complete — truncated URLs fail extraction.
• EC Press Corner is an Angular SPA — rag-web-browser returns 0 chars. Use REST API.

Full failure-mode catalog: reference/PIPELINE_DETAIL.md, reference/SOURCE_CONFIGS.md.

Reference

• data/sources.json — machine-readable: 33 sources × discovery × extractor × cost × gotchas
• reference/SOURCE_CONFIGS.md — per-source curl examples, run IDs, output samples
• reference/PIPELINE_DETAIL.md — RSS feeds, headline sources, warnings, failed paths, source quick-reference table
• reference/MACRO_SOURCES.md — ING Think, IMF, central banks
• reference/EUROPEAN_COMPANIES_GUIDE.md — European company handling tips
• reference/MORNING_NOTE_DETAIL.md — HTML output spec (archived; out of scope for this skill)
• reference/scripts/extract_and_clean.py, reference/scripts/clean_article.py — readability-lxml post-processors

European Companies Guide — Portfolio Focus

Shared reference for handling European companies across all analysis skills.

SEC Filing Availability

• Many European companies don't file with SEC (no CIK)
• ADR-listed European companies (e.g. AUTL) DO have SEC filings
• For non-SEC filers, look for annual reports on company investor relations pages
• For EU-listed companies, use Yahoo Finance and FMP as primary data sources

Reporting Frequency

• Quarterly reporting may be limited — semi-annual reporting is common for EU firms
• When quarterly data is unavailable, use semi-annual figures and interpolate cautiously
• Check company IR page for reporting calendar

Currency Handling

• Use EUR as base currency unless company reports in local currency
• Apply current FX rates from Yahoo Finance or Alpha Vantage
• Note currency of each data point in output tables
• Key portfolio currencies: CZK, PLN, SEK, HUF, BGN, TRY, GBP, USD

Data Source Priority (for non-SEC filers)

1. Company investor relations (annual/semi-annual reports) 2. Yahoo Finance (prices, basic financials) 3. Financial Modeling Prep (ratios, estimates) 4. Finnhub (real-time quotes, basic financials) 5. Alpha Vantage (FX rates, technical indicators)

Macro Economic Briefing Sources

For country-level economic context (daily/weekly macro overview), use these sources alongside news intelligence:

Best Open-Access Macro Sources

Central Banks (direct URL extraction, not via Google News)

Usage in Morning Briefing

When running morning briefing (v2), add a Market Context section: 1. Check think.ing.com for overnight CEE FX/rates commentary (query: site:think.ing.com Czech OR Poland OR Hungary) 2. Check ecb.europa.eu for any rate decisions or projections (especially around ECB meeting dates) 3. For specific country deep-dives, scrape the central bank URL directly

Morning Note Detail — Reference

Detailed morning note formatting, output schemas, and guidelines. Referenced from SKILL.md — load when producing portfolio morning briefing.

HTML Newsletter Format

Use Design System (see ../design-system/):

• Navy header with brand logo
• Dark title bar with date and "Top Call" highlight
• Collapsible company sections with logos and Portfolio/Competitor badges
• Copper accents for action items and trade ideas
• Separate Tier 1 / Tier 2 sections
• Fonts: Neue Haas Unica W1G (body), PPFDisplay (headings)
• Reference template: News_Intelligence_Report_2026-03-20.html in project root

JSON Output Schema

For downstream processing, output structured JSON:

{
  "morning_note": {
    "date": "2026-03-20",
    "top_call": {
      "headline": "...",
      "company": "InPost",
      "event_type": "earnings",
      "take": "...",
      "action": "Review PT"
    },
    "companies": [
      {
        "name": "InPost",
        "event_type": "earnings",
        "summary": "...",
        "take": "...",
        "action": "Review PT",
        "source": "bloomberg.com",
        "date": "2026-03-19"
      }
    ],
    "competitors": [...],
    "market_context": "...",
    "key_events_today": [...],
    "trade_ideas": [
      {
        "direction": "Long",
        "company": "InPost",
        "thesis": "...",
        "catalyst": "...",
        "risk": "..."
      }
    ],
    "no_material_news": ["Air Bank", "PPF Real Estate"]
  },
  "tier1": {
    "sources_queried": 8,
    "articles_found": 5,
    "articles": [
      {
        "tier": "verified",
        "title": "...",
        "source": "bloomberg.com",
        "author": "...",
        "published_date": "2026-03-19T14:17:09Z",
        "url": "...",
        "text_chars": 7577,
        "text_preview": "First 200 chars...",
        "discovery_method": "google_news_site",
        "extraction_actor": "jamie_tran/bloomberg-article-scraper",
        "extraction_cost": 0.02
      }
    ]
  },
  "tier2": {
    "articles_found": 3,
    "articles": [
      {
        "tier": "broad",
        "title": "...",
        "source": "seekingalpha.com",
        "url": "...",
        "text_chars": 1200,
        "extraction_actor": "apify/rag-web-browser",
        "extraction_cost": 0.005,
        "quality_warning": null
      }
    ]
  },
  "total_extraction_cost": 0.15
}

Earnings Analysis Template

When a portfolio company reports earnings, include this table:

Our Take: [Is this a beat or miss that matters? Quality of earnings? Guidance trajectory?] Action: [Maintain rating / Review PT / Flag for deep dive]

M&A Event Analysis Template

For M&A events, assess:

• Deal terms (premium to undisturbed price, structure, conditionality)
• Strategic fit for portfolio
• Probability of completion (regulatory, financing, shareholder approval)
• Impact on related portfolio companies

Important Notes

• Be opinionated — notes without a view are useless. If no view can be formed, state "insufficient data to form view, monitoring"
• Lead with the most important thing — Top Call is #1, do not bury the headline
• "No news" is valid — "Nothing material overnight across [N] portfolio companies. Maintaining positioning."
• Distinguish signal from noise — a minor analyst note is not the same as an earnings miss. Use priority classification.
• Time-stamp takes — if writing at 6am, note that pre-market may change by open
• Own mistakes — if yesterday's take was wrong, acknowledge it. Credibility > being right every time.
• Competitor context matters — a competitor's earnings miss can be bullish for portfolio company
• Don't over-extract — for morning note, extract only material articles (typically 3–10 per run). Skip general coverage unless slow news day.

Pipeline Detail — Reference

Detailed pipeline instructions, RSS feeds, headline sources, warnings, and failed paths. Referenced from SKILL.md — load when executing the pipeline.

RSS Discovery

Actor: louvre/rss-news-aggregator

Actor: louvre/rss-news-aggregator

Max 10 feeds per run — split into batches if needed.

14 sources with working RSS feeds:

No RSS: Reuters, Barron's, IBD, Les Echos, Caixin, Zawya, PAP (WAF), BTA, TASR, MTI (paywall).

curl -s -X POST "https://api.apify.com/v2/acts/louvre~rss-news-aggregator/runs?waitForFinish=120" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"rss_feeds":[{"url":"https://feeds.bloomberg.com/markets/news.rss"},{"url":"https://www.ft.com/rss/home"},{"url":"https://feeds.a.dj.com/rss/RSSMarketsMain.xml"},{"url":"https://www.economist.com/finance-and-economics/rss.xml"},{"url":"https://www.cnbc.com/id/100003114/device/rss/rss.html"},{"url":"https://www.forbes.com/business/feed2"},{"url":"https://www.morningstar.co.uk/uk/news/rss.aspx"},{"url":"https://www.marketwatch.com/rss/topstories"},{"url":"https://www.afr.com/rss/feed.xml"},{"url":"https://www.scmp.com/rss/91/feed"}],"raw_data":false}'

Batch 2 (remaining 4):

curl -s -X POST "https://api.apify.com/v2/acts/louvre~rss-news-aggregator/runs?waitForFinish=120" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"rss_feeds":[{"url":"https://asia.nikkei.com/rss/feed/nar"},{"url":"https://www.euronews.com/rss?level=theme&name=business"},{"url":"https://www.intellinews.com/rss/"},{"url":"https://www.handelsblatt.com/contentexport/feed/schlagzeilen"}],"raw_data":false}'

RSS returns general articles — filter client-side by company name/ticker match in title/description.

Headline Discovery

Actor: rodrigo_pacelli/headline-news-scraper

Actor: rodrigo_pacelli/headline-news-scraper

Only 6 sources work (others return 0):

curl -s -X POST "https://api.apify.com/v2/acts/rodrigo_pacelli~headline-news-scraper/runs?waitForFinish=120" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"urls":["https://www.cnbc.com","https://www.scmp.com","https://asia.nikkei.com","https://www.caixinglobal.com","https://www.handelsblatt.com"],"includeImages":false,"classifyWithAI":false}'

WARNING: Exclude Zawya from headline runs — returns category/tag pages, not article URLs.

No keyword filtering — filter results client-side by company name match.

Non-Article URL Filter Patterns

3. Non-article URL filter (apply to BOTH tiers): Discard URLs matching any of these patterns — they are NOT articles:

• Stock/quote pages: /markets/companies/, /quote/, /key-metrics/, /stock/
• Sitemaps & feeds: /sitemap, /rss/, /feed/
• Newsletters & redirects: /newsletters/, redirectUrl=
• Author/profile pages: /author/, /journalist/, /profile/
• Tag/category listing pages: /primarykeyword/, /topic/, /tag/ (except when part of article slug)
• Data pages with no article content: URL contains only ticker symbol and no article slug

Source Quick Reference Table

Critical Warnings

1. ALWAYS `decodeUrls: true` in Google News — encoded redirect URLs break ALL extractors 2. "InPost" is ambiguous — matches "post-Maduro", "post-ESG". Use "InPost SA" for precision 3. RSS max 10 feeds per run — must split into batches 4. Zawya headline-scraper returns category/tag pages (/primarykeyword/OIL), NOT article URLs — use Google News for Zawya discovery 5. Morningstar domain matters — .co.uk URLs fail with workhard3000. Use .com URLs only 6. WSJ livecoverage pages fail extraction — skip URLs matching wsj.com/livecoverage/ 7. Barron's/MarketWatch cost $0.25/page — use selectively for high-value articles only 8. Les Echos: Use region_language: "FR:fr" for French queries 9. Handelsblatt: Use region_language: "DE:de" for German queries 10. Caixin URLs must be complete — truncated URLs fail extraction 11. Paywall tightening observed — AFR (1,231 chars), SCMP (1,819 chars), Les Echos (2,832 chars) return less than earlier audits

Failed Paths (Do Not Use)

Output Format Detail

Structure extracted articles as JSON with separate tiers:

{
  "company": "InPost",
  "ticker": "INPST.AS",
  "query_date": "2026-03-19",
  "timeframe": "7d",
  "tier1": {
    "sources_queried": 6,
    "articles_found": 2,
    "articles": [
      {
        "tier": "verified",
        "title": "InPost Readies AI Shopping Assistant",
        "source": "bloomberg.com",
        "author": "Konrad Krasuski",
        "published_date": "2026-03-19T14:17:09Z",
        "url": "https://www.bloomberg.com/news/articles/...",
        "text_chars": 7577,
        "text_preview": "First 200 chars...",
        "discovery_method": "google_news_site",
        "extraction_actor": "jamie_tran/bloomberg-article-scraper",
        "extraction_cost": 0.02
      }
    ]
  },
  "tier2": {
    "articles_found": 5,
    "articles": [
      {
        "tier": "broad",
        "title": "InPost expands parcel locker network in France",
        "source": "seekingalpha.com",
        "url": "https://seekingalpha.com/news/...",
        "text_chars": 1200,
        "text_preview": "First 200 chars...",
        "discovery_method": "google_news_broad",
        "extraction_actor": "apify/rag-web-browser",
        "extraction_cost": 0.005,
        "quality_warning": null
      }
    ]
  },
  "total_extraction_cost": 0.15,
  "warnings": [],
  "failed_extractions": []
}

Present TWO separate tables to the user:

## News Intelligence: InPost (INPST.AS) — Last 7 days

Source Configurations — 33 Verified Financial News Sources + Macro Briefing

Each source includes verified MCPC commands, curl commands, input schemas, run IDs, and output samples. All configs tested 2026-03-19 with real Apify API runs. New sources (21-33) tested 2026-03-20.

Post-Processing: readability-lxml Cleanup

ALL `rag-web-browser` commands below use `outputFormats: ["html"]`. The raw HTML MUST be cleaned via readability-lxml before use — see Step 4c in SKILL.md for the clean_article() function.

This strips navigation, menus, language pickers, sidebars, and footers (30-70% noise reduction). Exceptions: Bloomberg scraper, workhard3000 extractor, and EC Press Corner API return already-clean output — no cleanup needed.

---

1. Bloomberg (bloomberg.com)

Discovery

Guidance:

• Hard paywall — Google News + RSS are the only discovery methods
• ALWAYS set decodeUrls: true — Bloomberg extractors fail on encoded Google News redirect URLs
• Use "InPost SA" not "InPost" to avoid false positives

Google News:

curl -X POST "https://api.apify.com/v2/acts/data_xplorer~google-news-scraper-fast/runs?waitForFinish=120" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"keywords":["site:bloomberg.com \"InPost SA\" OR \"INPST\""],"maxArticles":10,"timeframe":"7d","region_language":"US:en","decodeUrls":true,"extractImages":true,"proxyConfiguration":{"useApifyProxy":true,"apifyProxyGroups":["RESIDENTIAL"]}}'

• Verified: run TGlLVnklOTKiU4Ert — 30 articles (console)
• GNews verified: run 4GxdXYWLAFTbc9H3j — 1 article (console)

Extraction

Primary: `jamie_tran/bloomberg-article-scraper` — $0.02/article, 80 fields, structured body, terminal tickers

Command:

curl -X POST "https://api.apify.com/v2/acts/jamie_tran~bloomberg-article-scraper/runs?waitForFinish=120" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"start_urls":[{"url":"https://www.bloomberg.com/news/articles/2026-03-19/jpmorgan-goldman-offer-hedge-funds-way-to-short-private-credit"}]}'

• Verified: run qQbwWxgm0Ircjfm8h — 7,577 chars, 80 fields (console)
• Key fields: headline, authors, publishedAt, terminalTickers, body, byline, contentTags, aiSummary, url

Output sample:

{
  "headline": "JPMorgan, Goldman Offer Hedge Funds Way to Short Private Credit",
  "byline": "Silas Brown and Nishant Kumar",
  "publishedAt": "2026-03-19T14:17:09.116Z",
  "terminalTickers": [{"id": "GS:US"}, {"id": "CS:FP"}],
  "contentTags": [{"id": "private-credit", "name": "private credit", "type": "Topic"}]
}

Fallback 1: `romy/bloomberg-news-scraper` — $0.03/result, ~50% success rate, 38 fields

• Verified: run KMZxdgHg0YBi45hV4 — 7,040 chars (console)

Fallback 2: `workhard3000/news-intelligence-rag-extractor` — $0.025, 14 fields, last resort

• Verified: run o8GkLCESzmlW5380e — 2,996 chars (console)

---

2. Reuters (reuters.com)

Discovery

Guidance:

• Reuters deprecated ALL public RSS feeds — Google News is the only discovery method
• dadhalfdev/reuters-scraper-per-event IGNORES keyword param — do not use
• headline-news-scraper returns 0 for Reuters

Google News:

curl -X POST "https://api.apify.com/v2/acts/data_xplorer~google-news-scraper-fast/runs?waitForFinish=120" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"keywords":["site:reuters.com \"InPost SA\" OR \"INPST\""],"maxArticles":10,"timeframe":"7d","region_language":"US:en","decodeUrls":true,"extractImages":true,"proxyConfiguration":{"useApifyProxy":true,"apifyProxyGroups":["RESIDENTIAL"]}}'

Extraction

WARNING (updated 2026-03-24): Reuters blocks rag-web-browser completely — returns 0 chars on ALL URLs, even with RESIDENTIAL proxy. Use workhard3000 as primary extractor.

Primary: `workhard3000/news-intelligence-rag-extractor` with RESIDENTIAL proxy — $0.025

curl -X POST "https://api.apify.com/v2/acts/workhard3000~news-intelligence-rag-extractor/runs?waitForFinish=120" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"articleUrls":["https://www.reuters.com/..."],"autoArchive":true,"maxRetries":3,"requestIntervalMs":2000,"proxyConfiguration":{"useApifyProxy":true,"apifyProxyGroups":["RESIDENTIAL"]}}'

• Verified: 2026-03-24 — 2,632 chars with RESIDENTIAL proxy
• rag-web-browser returns 0 chars on ALL Reuters URLs (tested 5 URLs, standard + RESIDENTIAL proxy) — do NOT use
• Previous workhard3000 test without RESIDENTIAL (2026-03-19) returned only 386 chars — RESIDENTIAL proxy is critical

---

3. Financial Times (ft.com)

Discovery

Guidance:

• RSS (13 articles) + Google News
• headline-news-scraper returns articles but no keyword filtering — general FT front page

• RSS verified: run TGlLVnklOTKiU4Ert — 13 articles (console)

Extraction

Primary: `workhard3000/news-intelligence-rag-extractor` — $0.025/success (ONLY working extractor for FT)

Command:

curl -X POST "https://api.apify.com/v2/acts/workhard3000~news-intelligence-rag-extractor/runs?waitForFinish=120" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"articleUrls":["https://www.ft.com/content/..."],"autoArchive":true,"maxRetries":3,"requestIntervalMs":2000,"proxyConfiguration":{"useApifyProxy":true,"apifyProxyGroups":["RESIDENTIAL"]}}'

• Verified: run AlP9aKovUBRxhdJwc — 6,960 chars (console)
• rag-web-browser returns 16 chars ("Client Challenge") — FT blocks it

Output sample:

{
  "title": "ECB holds interest rates at 2% as energy prices soar",
  "byline": "Olaf Storbeck and Ian Smith",
  "publishedDate": "2026-03-19T13:15:18.282Z",
  "domain": "ft.com",
  "text": "The European Central Bank has kept its benchmark interest rate on hold at 2 per cent..."
}

---

4. WSJ (wsj.com)

Discovery

• RSS verified: run TGlLVnklOTKiU4Ert — 20 articles (console)

Extraction

Primary: `workhard3000/news-intelligence-rag-extractor` — $0.025/success (ONLY working extractor for WSJ)

• Verified: run 2zXshJwxycAKgN7yK — 4,748 chars (console)
• WARNING: byline and publishedDate are null — get from Google News/RSS discovery metadata
• WARNING: wsj.com/livecoverage/ URLs FAIL — skip them

---

5. Economist (economist.com)

Discovery

Guidance:

• Economist has EXCELLENT RSS — 300 articles! Best coverage of all 33 sources
• RSS should be primary, supplemented by Google News for targeted search

• RSS verified: run TGlLVnklOTKiU4Ert — 300 articles (console)

Extraction

Primary: `workhard3000/news-intelligence-rag-extractor` — $0.025/success (ONLY working extractor)

• Verified: run ckhijvIQcwvwaOdaw — 14,256 chars (console)
• Byline always "The Economist" — no individual authors
• rag-web-browser returns 0 chars for Economist

---

6. CNBC (cnbc.com)

Discovery

3 working methods: RSS (30), Headlines (19), Google News.

• RSS verified: run TGlLVnklOTKiU4Ert — 30 articles
• Headlines verified: run a6JHWQ6LCv8DCum13 — 19 articles (console)

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: run 2QKQRPjZRyJxfggJt — 4,865 chars (console)
• Optional: workhard3000 for metadata (byline, date, image) — run 5WBPe82sgy5rUsigG, 2,837 chars

---

7. Forbes (forbes.com)

Discovery

RSS (25 articles) + Google News.

• RSS verified: run TGlLVnklOTKiU4Ert — 25 articles

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: run naBDjdL8RHZK4MEKM — 4,969 chars (console)
• workhard3000 returns 0 chars for Forbes — do NOT use
• natasha.lekh Forbes actor is UNDER MAINTENANCE — do not use

---

8. Morningstar (morningstar.com)

Discovery

RSS (.co.uk, 20 articles) + Google News (.com).

IMPORTANT: morningstar.co.uk URLs fail with workhard3000 — use morningstar.com URLs for extraction.

• RSS verified: run TGlLVnklOTKiU4Ert — 20 articles

Extraction

Primary: `workhard3000/news-intelligence-rag-extractor` — $0.025/success

• Verified: run uVI3NxwXu9rULA5zY — 35,256 chars (console)
• janbruinier Morningstar actor is for fund ratings, NOT news — do not use

---

9. Barron's (barrons.com)

Discovery

ONLY Google News — RSS returns 403, headline-scraper returns 0.

Extraction

Primary: `stanvanrooy6/universal-ai-web-scraper` — ~$0.25/page (USE SELECTIVELY)

Command:

curl -X POST "https://api.apify.com/v2/acts/stanvanrooy6~universal-ai-web-scraper/runs?waitForFinish=120" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"startUrls":[{"url":"https://www.barrons.com/articles/..."}],"instructions":"Extract the full article text, title, author name, publication date, and any tags or categories.","proxyConfiguration":{"useApifyProxy":true,"apifyProxyGroups":["RESIDENTIAL"]},"useWebSearch":true}'

• Verified: run tKoc6e4qWFGhHYG2x — 3,387 chars (console)
• Output fields: author, categories, fullArticleText, publicationDate, tags, title

Output sample:

{
  "title": "3 Reasons This Drone Stock Soared 520%...",
  "author": "Al Root",
  "publicationDate": "2026-03-18T10:38:00Z",
  "categories": ["daily", "feature"],
  "tags": ["Aerospace and Defense", "Drones", "IPO"]
}

---

10. MarketWatch (marketwatch.com)

Discovery

Google News (primary) + RSS (10 articles, small volume).

• RSS verified: run TGlLVnklOTKiU4Ert — 10 articles

Extraction

Primary: `stanvanrooy6/universal-ai-web-scraper` — ~$0.25/page (USE SELECTIVELY)

• Verified: run E41lckeBGwYoFQMMl — fullArticleText present (console)
• rag-web-browser returns HTTP 500 — completely blocked

Cheap fallback (excerpts only): workhard3000 — $0.025, returns 448-878 chars

• Verified: run FdKa2X4C8IyQYTTck — 559 chars (console)

---

11. IBD / Investor's Business Daily (investors.com)

Discovery

Google News only (very limited for non-US companies). RSS returns 403.

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: run a9dwfryRba07Eb0Nh — 6,722 chars (console)
• Optional metadata: workhard3000 run PHQtaoYk0xa1f2QjH — 958 chars (byline, date, image)

---

12. Les Echos (lesechos.fr)

Discovery

ONLY Google News — RSS returns 403. Use French queries with region_language: "FR:fr".

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: run K4sLa7tbZ720xwAkq — 2,832 chars (console)
• Paywall tightened: was 9,257 chars in audit, now 2,832
• Content in French
• Optional metadata: workhard3000 run PHQtaoYk0xa1f2QjH — 716 chars (byline, date)

---

13. AFR / Australian Financial Review (afr.com)

Discovery

RSS (20 articles) + Google News. AFR focuses on APAC — limited European coverage.

• RSS verified: run C8Q7CeLmLwEKbGLLS — 20 articles (console)

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: run Mod5b0kwyjdpR7Nea — 1,231 chars (console)
• WARNING: Paywall tightened — was 4,399 chars, now only 1,231

---

14. SCMP / South China Morning Post (scmp.com)

Discovery

Excellent discovery: RSS (50) + Headlines (39) + Google News.

• RSS verified: run C8Q7CeLmLwEKbGLLS — 50 articles
• Headlines verified: run a6JHWQ6LCv8DCum13 — 39 articles

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: run LYVVUSk197JTJu81z — 1,819 chars (console)
• WARNING: Paywall tightened — was 3,014 chars, now 1,819

---

15. Nikkei Asia (asia.nikkei.com)

Discovery

Best headline coverage (60!) + RSS (50, no description) + Google News.

• Headlines verified: run a6JHWQ6LCv8DCum13 — 60 articles (highest!)
• RSS verified: run C8Q7CeLmLwEKbGLLS — 50 articles (title + link only, no description)

Extraction

Primary: `workhard3000/news-intelligence-rag-extractor` — $0.025/success (outperforms rag-web-browser for Nikkei)

• Verified: run XK7gMVH4Ddzk0B7YF — 5,723 chars (console)
• Full metadata: byline ("EISAKU NITTA"), publishedDate, image

---

16. Caixin Global (caixinglobal.com)

Discovery

Google News + Headlines (7, very low volume). No RSS.

• Headlines verified: run a6JHWQ6LCv8DCum13 — 7 articles

Extraction

Primary: `workhard3000/news-intelligence-rag-extractor` — $0.025/success

• Verified: run 2oYfpd4YhrSQc50Lg — 4,166 chars (console)
• Free/commentary = full text; paywalled news = ~885 chars
• IMPORTANT: Use complete URLs — truncated URLs fail
• rag-web-browser fetches wrong page for Caixin — do NOT use

---

17. Zawya (zawya.com)

Discovery

Google News is recommended. Headlines (50 articles) exist but return tag pages, not articles.

WARNING: Headline scraper returns category/tag page URLs (e.g. /primarykeyword/OIL), NOT article URLs. Filter these out or avoid.

Extraction

Primary: `workhard3000/news-intelligence-rag-extractor` — $0.025/success

• Verified: run RzRvX08Xg29SsrXV5 — 12,681 chars (console)
• rag-web-browser returns HTTP 500 for Zawya — do NOT use
• Full metadata: byline, publishedDate, image

---

18. Euronews (euronews.com)

Discovery

RSS (50 articles, business theme) + Google News. NOT paywalled.

• RSS verified: run C8Q7CeLmLwEKbGLLS — 50 articles

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: run fREgpuCzO6DulRH8w — 10,917 chars (console)
• Optional metadata: workhard3000 run 5WBPe82sgy5rUsigG — 8,670 chars (byline: "Denis Loctier", date, image)

---

19. IntelliNews (intellinews.com)

Discovery

RSS (variable volume) + Google News. Niche source, excellent for CEE coverage.

• RSS verified: run C8Q7CeLmLwEKbGLLS — volume varies (console)

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: run eAiUialgv1waTvJNP — 10,744 chars (console)
• Optional metadata: workhard3000 run 5WBPe82sgy5rUsigG — 6,496 chars
• NOTE: publishedDate is null in workhard3000 — get from RSS/GNews discovery

---

20. Handelsblatt (handelsblatt.com)

Discovery

Excellent: Headlines (60) + RSS (50, full metadata) + Google News. German language.

• Headlines verified: run a6JHWQ6LCv8DCum13 — 60 articles
• RSS verified: run C8Q7CeLmLwEKbGLLS — 50 articles with author + date

Google News — MCPC (German):

curl -s -X POST "https://api.apify.com/v2/acts/data_xplorer~google-news-scraper-fast/runs?waitForFinish=120" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"keywords":["site:handelsblatt.com \"InPost SA\" OR \"INPST\" OR \"Paketautomaten\""],"maxArticles":10,"timeframe":"7d","region_language":"DE:de","decodeUrls":true,"extractImages":true,"proxyConfiguration":{"useApifyProxy":true,"apifyProxyGroups":["RESIDENTIAL"]}}'

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: run Y79uVeyDxC0dD6ncY — 10,435 chars (console)
• Content in German
• Optional metadata: workhard3000 run PHQtaoYk0xa1f2QjH — 7,737 chars (byline: "Laura de la Motte", date, image)

---

Pan-European Sources — Added 2026-03-20

---

21. POLITICO Europe (politico.eu)

Discovery

• Verified: 2026-03-20, 5 articles — strong EU policy/energy coverage

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

curl -s -X POST "https://api.apify.com/v2/acts/apify~rag-web-browser/runs?waitForFinish=120" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"query":"<POLITICO_ARTICLE_URL>","maxResults":1,"outputFormats":["html"],"requestTimeoutSecs":40,"proxyConfiguration":{"useApifyProxy":true,"apifyProxyGroups":["RESIDENTIAL"]},"removeCookieWarnings":true}'

• Verified: 2026-03-20 — 7,677 chars, full article text for free articles
• Premium/paywalled articles may return partial content — still useful for headline + lead

---

22. EUobserver (euobserver.com)

Discovery

• Verified: 2026-03-20, 5 articles — investigative/transparency focus

Extraction

Primary: `workhard3000/news-intelligence-rag-extractor` (with autoArchive) — $0.025

curl -s -X POST "https://api.apify.com/v2/acts/workhard3000~news-intelligence-rag-extractor/runs?waitForFinish=120" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"articleUrls":["<EUOBSERVER_URL>"],"autoArchive":true,"maxRetries":3,"requestIntervalMs":2000,"proxyConfiguration":{"useApifyProxy":true,"apifyProxyGroups":["RESIDENTIAL"]}}'

• Verified: 2026-03-20 — 3,001 chars via archive bypass
• rag-web-browser hits "Unlock article" paywall — do NOT use as primary

---

23. EUbusiness (eubusiness.com)

Discovery

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: 2026-03-20 — 4,828 chars, open access, clean extraction

---

24. EU Reporter (eureporter.co)

Discovery

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: 2026-03-20 — 5,921 chars, open access, full article text

---

Institutional Sources — Added 2026-03-20

---

25. EC Press Corner (ec.europa.eu)

Discovery

Extraction

DO NOT use rag-web-browser — EC Press Corner is an Angular SPA, JS-only render returns 0 chars.

Use REST API directly:

curl -s "https://ec.europa.eu/commission/presscorner/api/documents?reference=IP/26/614&language=en" | python3 -c "
import sys,json
d=json.load(sys.stdin)
html=d.get('docuLanguageResource',{}).get('htmlContent','')
print(f'Chars: {len(html)}')
# Strip HTML tags for plain text
import re
text=re.sub('<[^<]+?>','',html)
print(text[:500])
"

• Verified: 2026-03-20 — REST API returns 9K+ chars clean HTML per press release
• Parse IP_XX_NNN reference IDs from Google News titles to construct API calls
• Subdomain pages (energy.ec.europa.eu, research.ec.europa.eu) extract fine with rag-web-browser (1,936 chars)

---

26. ECB (ecb.europa.eu)

Discovery

• Verified: 2026-03-20, 5 articles — projections, rate decisions, digital euro

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: 2026-03-20 — press releases: 6,950 chars; projection reports: 85,009 chars
• First ~400 chars are nav boilerplate — strip or ignore
• Excellent institutional source for SK and BG (eurozone members)

---

CEE Local Sources — Added 2026-03-20

---

27. ČTK / České noviny (ceskenoviny.cz)

Discovery

• GNews verified: 2026-03-20 — 5 articles (CZ:cs region)
• RSS verified: 2026-03-20 — 50 articles per feed (ekonomika.php, cr.php)
• Search on ceskenoviny.cz does NOT work (JS rendered, returns shell only)

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: 2026-03-20 — 4,596 chars, full article text in Czech

---

28. PAP — Polska Agencja Prasowa (pap.pl)

Discovery

• Verified: 2026-03-20 — 5 articles
• PAP web search returns false positives (3-letter string match)
• PAP RSS is blocked by WAF (Incapsula)
• Only Google News site:pap.pl works reliably

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: 2026-03-20 — 9,232 chars, full article text in Polish

---

29. BTA — Bulgarska Telegrafna Agentsia (bta.bg)

Discovery

• Verified: 2026-03-20 — 5 articles
• BTA web search is JS rendered — returns empty results
• BTA has no RSS feeds
• Only Google News site:bta.bg works

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: 2026-03-20 — 10,851 chars, full article text in Bulgarian

---

30. TASR — Tlačová agentúra Slovenskej republiky (tasr.sk)

Discovery

• Verified: 2026-03-20 — 5 articles
• WARNING: Google News returns generic titles ("Tlačová agentúra Slovenskej republiky") — extraction needed to determine article content
• TASR web search redirects to homepage (SPA)
• TASR has no RSS feeds

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: 2026-03-20 — 3,956 chars, article text in Slovak

---

31. Telex.hu (telex.hu) — Hungary

Discovery

• Verified: 2026-03-20 — 2 articles for portfolio-related queries
• Independent outlet, launched by ex-index.hu journalists — high editorial credibility
• Best Hungarian extraction quality among tested sources

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: 2026-03-20 — 8,216 chars, excellent clean article body in Hungarian

---

32. HVG.hu (hvg.hu) — Hungary

Discovery

• Verified: 2026-03-20 — 3 articles
• Respected independent weekly (since 1968), strong digital presence

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: 2026-03-20 — 4,883 chars, clean article body in Hungarian

---

33. Világgazdaság (vg.hu) — Hungary

Discovery

• Verified: 2026-03-20 — 4 articles (strongest GNews presence among HU sources)
• Hungarian business daily — focuses on economics and finance

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: 2026-03-20 — 9,906 chars, article body in Hungarian
• First ~900 chars contain navigation/ticker noise — article body starts after

---

Macro Briefing Sources (not Tier 1, but used for market context)

---

ING Think (think.ing.com)

Discovery

curl -s -X POST "https://api.apify.com/v2/acts/data_xplorer~google-news-scraper-fast/runs?waitForFinish=120" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"keywords":["site:think.ing.com Czech OR Poland OR Hungary OR CEE"],"maxArticles":10,"timeframe":"7d","region_language":"US:en","decodeUrls":true,"proxyConfiguration":{"useApifyProxy":true,"apifyProxyGroups":["RESIDENTIAL"]}}'

Extraction

Primary: `apify/rag-web-browser` — ~$0.005/query

• Verified: 2026-03-20 — 11,319 chars, excellent quality
• Best free open-access source for daily CEE FX, rates, and macro commentary
• Covers CZ, PL, HU, SK, BG + broader EU

---

IMF Country Pages (imf.org)

Extraction

curl -s -X POST "https://api.apify.com/v2/acts/apify~rag-web-browser/runs?waitForFinish=120" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"query":"https://www.imf.org/en/Countries/CZE","maxResults":1,"outputFormats":["html"],"requestTimeoutSecs":40,"proxyConfiguration":{"useApifyProxy":true,"apifyProxyGroups":["RESIDENTIAL"]},"removeCookieWarnings":true}'

• Verified: 2026-03-20 — 22,250 chars for Article IV concluding statements
• Annual cadence (Article IV) + quarterly (WEO data)
• Country codes: CZE (CZ), POL (PL), HUN (HU), BGR (BG), SVK (SK)

---

ČNB — Czech National Bank (cnb.cz)

Direct URL Extraction

curl -s -X POST "https://api.apify.com/v2/acts/apify~rag-web-browser/runs?waitForFinish=120" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"query":"https://www.cnb.cz/en/monetary-policy/","maxResults":1,"outputFormats":["html"],"requestTimeoutSecs":40,"proxyConfiguration":{"useApifyProxy":true,"apifyProxyGroups":["RESIDENTIAL"]},"removeCookieWarnings":true}'

• Verified: 2026-03-20 — 35,367 chars (full bank board decisions listing)
• WARNING: Google News site:cnb.cz is blocked by Cloudflare — use direct URL extraction only
• Use keyword query "ČNB" or "Czech National Bank" in broad GNews search instead