Skip to main content
Structured, company-attributed news from the articles Specter monitors. Each signal is one classified company mention within a single article (so an article naming three companies produces three signals), with an AI summary of what’s newsworthy and a 0–5 importance score.

Data categories

  • Article. article_id (shared across signals from the same article), article_url, title, published_date, featured_image_url.
  • Classification. meaningful_updates: REVENUE_UPDATE, PROFIT_UPDATE, TRACTION_UPDATE, FUNDING_UPDATE, DEAL_UPDATE.
  • Summary & importance. whats_new (AI summary), score (0–5; 4+ is major news), traction.
  • Company link. entity_id, company_name, company_domain.
Example (trimmed)

Endpoints

How to use the API

Start from a company News is served per company by Get company news, which returns that company’s complete news history from its Specter company ID. Only have a domain, website, or name? Resolve it first with Find company ID (free, up to 50 at once) or Search entities by name. Cover a list or a saved search There is no list- or search-scoped news endpoint; fan out over the members instead. Filter
  • Set major_only=true to restrict to signals scoring ≥ 4.
  • Bound the window with news_publication_date_after / news_publication_date_before (both inclusive) — the cheapest way to poll a large cohort is a narrow date range per company.
Expand from a signal
  • Company → follow entity_id to Get company by ID.
  • Article → every signal already carries the full article context (article_url, title, whats_new, score), so no follow-up lookup is needed.

Notes

  • One signal = one company mention in one article. Articles naming multiple companies share an article_id.
  • score runs 0–5; 4+ is treated as major news (major_only filters to those).
  • meaningful_updates classifies the news: revenue, profitability, traction, funding, or deals.
  • signal_id is a plain integer (not a prefixed string).
  • Get company news is charged per block of results returned rather than per signal, so a full page costs the same as a nearly empty one; pull wide pages instead of many narrow ones. See its “Costs” note for the figure.
  • An unknown company returns 404; a known company with no coverage returns 200 with an empty array.