Every relevant headline globally, tagged to ticker(s), graded for sentiment + novelty + materiality, joined to actual price impact (1m / 5m / 1h / 1d returns vs sector + index baseline). For quant funds, hedge funds, prop desks, financial-news LLM training. NDJSON+Parquet streaming or batch. Pricing on request — send your ticker universe + use-case, quote in 24h.
Continuous stream of headlines from 1,800 sources globally — newswires (Reuters, AP, Bloomberg wire), filings (SEC EDGAR, EU ESMA, UK FCA), exchanges, broker research syndication, financial press, and crypto/commodity-specific feeds. Each event is tickerized, sentiment-graded, novelty-scored against the rolling corpus, then joined to the actual 1m/5m/1h/1d price return to compute realized impact.
Designed for systematic strategies: pre-news baseline + multiple-horizon returns + sector/index relative. LLM sentiment is the headline + first 500 chars + source bias correction, graded by Claude Sonnet 4.6 + verified by a smaller distilled model for stability.
Real events. Five sample rows from the live stream. Full 24-hour backfill free for any one ticker — just send the symbol.
# NVDA · Q2 guidance raise · wire-first {"event_id":"evt_8a92f04c","ts_utc":"2026-05-20T13:42:08Z","source":"Reuters Business", "headline":"Nvidia raises Q2 revenue guidance by 8% on AI datacenter demand", "primary_ticker":"NVDA","tickers":["NVDA","AMD","AVGO"],"sector_gics":"semis", "sentiment":0.84,"novelty":0.91,"materiality":"high","topics":["guidance","earnings"], "price_t0":142.18,"return_1m":0.0142,"return_5m":0.0287,"return_1h":0.0319, "return_vs_sector":0.0214,"impact_score":87} # ASML · EU export-control extension · regulator-first {"event_id":"evt_c1b8e421","ts_utc":"2026-05-19T07:01:32Z","source":"EU Commission", "headline":"EU extends China DUV chip-export restrictions to 2028", "primary_ticker":"ASML.AS","tickers":["ASML.AS","NVDA","AMAT","LRCX"], "sentiment":-0.62,"novelty":0.78,"materiality":"high","topics":["regulation","trade"], "price_t0":684.50,"return_5m":-0.0218,"return_1h":-0.0341,"return_vs_sector":-0.0190} # TSLA · cybertruck recall · low-novelty (5th source to report) {"event_id":"evt_3f72a118","ts_utc":"2026-05-19T14:18:09Z","source":"Bloomberg", "headline":"Tesla recalls 47,000 Cybertrucks over windshield-wiper issue", "primary_ticker":"TSLA","sentiment":-0.34,"novelty":0.18,"materiality":"low", "duplicates_count":23,"first_source":"NHTSA filing","return_5m":-0.0012,"impact_score":14} # BTC · ETF inflow record · crypto {"event_id":"evt_e482001a","ts_utc":"2026-05-18T20:01:00Z","source":"CoinDesk", "headline":"Spot BTC ETFs see record $1.4B single-day net inflow", "primary_ticker":"BTC-USD","tickers":["BTC-USD","IBIT","FBTC"],"asset_class":"crypto", "sentiment":0.71,"return_1h":0.0184,"return_1d":0.0312,"impact_score":72} # SAP · DACH-only press, low US impact {"event_id":"evt_b18402d9","ts_utc":"2026-05-17T11:22:48Z","source":"Handelsblatt", "headline":"SAP gewinnt Lufthansa als Großkunden für RISE-Plattform","lang":"de", "primary_ticker":"SAP.DE","sentiment":0.58,"novelty":0.62,"materiality":"med", "return_1h":0.0089,"return_vs_sector":0.0054,"impact_score":41}
Backfill snapshot for backtesting. Streaming for live execution. Pricing is per use-case — ticker universe size, refresh cadence, latency SLA, license scope. Send the spec, we send a quote in 24h.
Full historical backfill since 2024-01-01 for one ticker universe (e.g. all S&P 500, or top-100 EU large-caps, or top-200 crypto). For strategy research, factor construction, signal R&D.
Real-time WebSocket feed for up to 5,000 tickers of your choice. For quant funds, prop desks, signal-driven execution. Plus rolling 90-day backfill.
All 12K+ tickers · full history since 2024-01 · live stream · full schema · dedicated support. For hedge funds and financial-news LLM-training programs.
Three things to know before you wire alpha to it.
Sentiment is not a finance-tuned bespoke model — it's Claude Sonnet 4.6 with a 12-shot prompt + a smaller distilled cross-check. Quality benchmarks ship in the README. If you want your own model, the raw headline + URL is in every event, swap it in.
Returns are computed from real consolidated tape data, not modeled. T0 = headline publication ts in UTC, T+N = mid-quote N seconds later. For markets closed at publication time, returns are computed from next open with a flag. Crypto uses 24h venues directly.
This is an aggregator — we listen, you act. Median 2.4s pipeline latency from publication to enriched event. If you need sub-second from a primary source, license Reuters/Bloomberg directly. We work as the cheaper second layer for sentiment + impact analytics.
Email us the ticker universe + tier + use-case. We send the sample + a written quote within 24h. Free full-history backfill for one ticker on first contact.
Request Free Sample →