Skip to main content
StockContext’s public claim is narrow and checkable: analyst-useful SEC facts for active US-listed companies, cited to the accession, fresh in minutes on the filing-driven path where applicable, and explicit reasons when a fact cannot be served. This page is that claim, enumerated. Two things make the catalog different from a source list. Every served value carries per-value provenance (the SEC accession, form, and filing date it came from — provenance=full on /v2/facts adds the exact XBRL tag), and every boundary is labeled in the payload itself: coverage eras, publication lags, and refusals are machine-readable fields, never footnotes. When a parse is uncertain, the API refuses with a reason instead of serving a guess.

What is collected and where it serves

DatasetServes onVerified by
Financial statements + per-value provenance, point-in-time history, restatement flags, IFRS/20-F, multi-class EPS/v2/fundamentals, /v2/facts, /v2/historynightly coherence referee (G-COHERE), external anchors vs re-parsed filings, golden values
Generic fact catalog (~150 concepts), as-known-on-date queries/v2/facts (as_of, provenance=full)G-PIT property gate; registry pinned to the engine catalog in CI
Filings, section text, section diffs/v2/filings (+ subresources)citation gate (G-CITE)
Classified events: 8-K items, late filings, delisting, deregistration, offerings/dilution, M&A/tender/going-private, comment letters, going-concern detections, 5% stakes/v2/events (frozen label vocabulary, labels= filter)classification goldens — every label pinned to ≥3 real filings in CI
Insider Form 4 transactions + distinct-buyer clusters + 10b5-1 flags + Form 144 intent-to-sell notices (Forms 3/5 are source-available future collection, not currently served)/v2/insidergoldens cross-checked against raw filing XML
Company-stated guidance (the estimates substitute)/v2/earnings.company_guidanceserve-or-refuse anchoring (every served number’s period/metric/unit/basis is text-anchored); 20-filing goldens; nightly parse-rate drift gate (G-GUID)
Auditor identity (name, PCAOB ID, location), filer category, public float/v2/profilethree hand-verified covers; float equal to the dollar vs the companyfacts API
ETF holdings (N-PORT, top-10) for the 57-ETF tier/v2/profile (ETF variant)three-fund goldens with offline replay (G-NPORT)
Fails-to-deliver (semi-monthly CNS)/v2/price-action?blocks=fails_to_deliverper-file row-count + date-window integrity; cycle-staleness refusal
Freshness is measured, not claimed: issuer-keyed filings flow filing → served in minutes (per-form receipts are published on the ops surface), and everything slower says so in its own payload — 13F carries as_of_quarter_end + filed_through, N-PORT carries as_of with its ~60-day publication lag, fails-to-deliver carries as_of_settlement_date + data_lag_days.

The estimates question

There is no analyst consensus here, ever — consensus data exists only inside commercial aggregators (LSEG, FactSet, Zacks) under licenses that are a legal trap for agent products that redistribute outputs. The substitutes are structurally better for agents and 100% redistributable:
  • what the company itself guided (company_guidance, extracted under a strict serve-or-refuse rule),
  • what the market price implies (/v2/priced-in reverse-DCF), and
  • what insiders and SEC event filings disclose (/v2/insider and /v2/events).

Private beta, hidden from public core

These families remain internally callable for explicitly enabled integrations, but are not in public MCP tools, endpoint chooser guidance, or default agent recipes until coverage gates are strong enough:
FamilyWhy it is private beta
SegmentsValuable for business-line analysis, but the coverage gate must distinguish issuer disclosure reality, taxonomy variation, and recast history before public-core positioning.
Ownership / 13F / 13D-GValuable for institutional and 5% holder analysis, but quarterly lag, class-level security matching, and completeness gates need more production sampling before public-core positioning.
Governance / proxy factsValuable for compensation and voting analysis, but proxy taxonomy coverage and collection completeness are still being validated.

What is excluded, and why (the honest perimeter)

ExcludedWritten reason
Analyst estimates / consensus / surprise-vs-consensusLicensed-only data; the razor (facts, math, provenance — never opinions). See the substitutes above.
NewsIntentionally absent forever — narrative is not a fact an agent can cite.
Delisted companies, OTC, pre-2009 backfillOwner scope decision: active US-exchange-listed only.
Fund proxy voting (N-PX)Fund-side data; issuer-side vote outcomes already serve via 8-K Item 5.07 events.
SEC enforcement / litigation releasesNot EDGAR; respondent→ticker matching is error-prone and fails refuse-on-uncertain economics.
Exchange short interestFINRA/exchange licensed data, not SEC. Fails-to-deliver is the SEC-side cousin and is served.
CEO/CFO certifications (EX-31/32)Boilerplate; the signal cases surface via late-filing, 4.02, and auditor events.
Proxy text beyond the ecd taxonomy (director bios, pre-2022 comp)No structured taxonomy exists — extraction would be scrape-grade, below the precision bar. Competitors’ deeper comp history is scrape-sourced.
Form D / Reg A / Form C / ADV / N-CEN / N-MFPPrivate-company / adviser / fund-ops surfaces outside issuer scope.
SPY’s N-PORT holdingsNot a gap: SPY is a unit investment trust, which the SEC does not require to file N-PORT. The payload says fund_files_no_nport.
Known boundaries inside public-core families are labeled in the payloads: auditor tags start FY2021; going-concern detection runs from 2026-06-12 forward (historical backfill queued); guidance covers the trailing ~120 days plus everything filed since 2026-06-12.