KeyVex

Returns OSHA workplace-safety enforcement records from the Department of Labor's enforcement data: inspection cases (who was inspected, where, why, when) with optional violation citations attached (standard cited, violation type, penalties, abatement dates). Use this when the user asks about: a company's workplace-safety record, OSHA penalties or citations, fatality/catastrophe investigations, inspection activity by state or industry (NAICS), or contractor safety research. Result rows are INSPECTIONS. Pass include_violations=true to attach each inspection's citations under a `violations` array (or pass activity_nr for a direct lookup, which always includes them). min_penalty keeps only inspections with at least one violation whose initial or current penalty meets the threshold (implies include_violations). insp_type codes (DOL's own legend): A=Accident, B=Complaint, C=Referral, D=Monitoring, E=Variance, F=FollowUp, G=Unprog Rel, H=Planned, I=Prog Related, J=Unprog Other, K=Prog Other, L=Other-L, M=Fat/Cat (fatality/catastrophe), N=Unprog Emph. Each record carries insp_type_label with the decoded value. Violation viol_type codes: S=Serious, W=Willful, R=Repeat, O=Other, U=Unclassified. Violations carry delete_flag='X' when the source later deleted the citation — rows are kept with the flag, never dropped. Filter combinations note: server-side indexes support ONE of state / naics_code combined with the open_date sort + since/until. Other filters (establishment_name substring, insp_type, state+naics together, min_penalty) post-filter client-side over a widened fetch window. sort_by=case_mod_date is the 'recently updated cases' firehose and cannot be combined with state / naics_code / since / until. Pure-publisher posture: DOL's enforcement rows as published — codes kept verbatim (with the source's own legend decoded alongside), no safety scoring. Each record's source_url links to the osha.gov establishment inspection-detail page.

A ticker search also returns rows this issuer FILED UNDER SYMBOLS IT NO LONGER LISTS (renames, filer typos, ADR spellings). Each row keeps `ticker` as SEC received it and gains `current_ticker` when the issuer trades under a different symbol today. Separately-listed share classes are NOT merged, and asking for a RETIRED symbol returns only rows filed under it — retired symbols get reissued to other companies. Returns Form 144 filings — notices of proposed sale by corporate insiders (officers, directors, 10%+ holders) under Rule 144 of the Securities Act. Each record is one planned-sale line from one filing. ⚠ aggregate_market_value is NULLABLE. A Form 144 that did not state a value now reports null rather than 0 — but a filer who genuinely stated 0.00 still reports 0, and that happens. Null never satisfies min_value and sorts last. ⚠ AND SOME FILERS STATE THE ISSUER'S MARKET CAP IN THAT BOX, WHICH PUTS THEM AT THE TOP OF A DESCENDING VALUE SORT. Measured 2026-09-04: 5 of the top 100 — SYF 4,000 shares stating $25.24bn ($6.31m per share), IT 860 shares stating $11.72bn. Dividing those by shares_outstanding on the same row gives $77.57 and $185.66, which are the real share prices. ⚠ THE CHECK CATCHES TWO OF THOSE FIVE, NOT ALL FIVE, AND SAYS SO RATHER THAN OVERSTATING ITSELF. TCMD, EPSM and XHR imply $36,017, $17,577 and $16,686 per share — absurd for those issuers, but below BRK.A's real $740,000 peak, so no per-row test separates them from a genuine high-priced sale. They still read 'checked'. Settling them needs a comparison against the day's actual price bar. Every row therefore carries aggregate_market_value_check: 'checked' the value implies a plausible price for the shares sold 'not_stated' no value was filed (distinct from a filed 0.00) 'no_share_count' no share count, so the check could not run 'implausible_looks_like_market_cap' implies a per-share price above any that has traded, and shares_outstanding yields a plausible one instead 'implausible_unexplained' implies an impossible price and shares_outstanding does not explain it — we can say it is not the sale value without being able to say what it is ⚠ The number is NOT corrected. SEC's bytes are served exactly as filed, and we do not invent a value the filer never stated. If you rank by this field, exclude anything whose check does not read 'checked' — otherwise the top of your list is market capitalisations. Use this when the user asks about: insiders who have announced they're about to sell, upcoming insider sales at a specific company, large planned sales by value, or which executives are signaling intent to exit positions. Form 144 is a *forward-looking* signal. It's filed BEFORE the actual sale, which later lands as a Form 4. The complement to get_insider_transactions: that tool tells you what insiders just did, this one tells you what they're about to do. Filing thresholds: ≥5,000 shares OR ≥$50,000 aggregate value. The aggregate_market_value is the insider's estimate at filing time; the actual sale price/value can differ. The approximate_sale_date is also an estimate — the real Form 4 transaction_date may be days later. Most Form 144 filings list one security line, but a single filing can cover multiple share classes (e.g., separate Class A + Class B). Each line is returned as its own record.

Returns SEC Form D filings — Reg D / Rule 506 private placement offering notices. Use this when the user asks about: who's raising private capital right now, new VC fund formations, private equity raises, real-estate syndicates, hedge fund launches, who's claiming Rule 506(b) vs 506(c) exemption, or to identify directors / executive officers of newly-formed entities. ⚠ total_amount_sold, min_investment_accepted, total_number_already_invested, sales_commissions and finder_fees are NULLABLE — a Form D that did not state a figure reports null rather than 0. The distinction matters here more than anywhere: a Form D filed at the START of an offering legitimately reports $0 sold, so 0 and null mean genuinely different things. Rows stored before 2026-08-17 cannot tell you which they were. Null never satisfies min_amount_sold and sorts last. Source: SEC EDGAR full-text search + per-filing primary_doc.xml. All Reg D filings (504 / 506(b) / 506(c)) plus Section 4(a) exempt offerings flow through here. Form D must be filed within 15 days of the first sale. Each record carries: issuer entity (name, CIK, address, jurisdiction of incorporation, entity type), offering data (industry group, investment fund type for pooled funds, total offering / sold / remaining, minimum investment, federal_exemptions claimed), filing metadata (file_date, date_of_first_sale, is_amendment), and a related_persons[] array of directors / executive officers / promoters. Federal exemption codes (federal_exemptions array): 06b — Rule 506(b) (no general solicitation; up to 35 non-accredited) 06c — Rule 506(c) (general solicitation OK; all accredited) 04(2) — Section 4(a)(2) (statutory private placement) 3C — ICA Section 3(c) (3(c)(1), 3(c)(5), 3(c)(7) etc. — common for funds) 3C.1 — ICA 3(c)(1) (up to 100 investors) 3C.7 — ICA 3(c)(7) (qualified purchasers only) Common industry_group_type values: 'Pooled Investment Fund' (with investment_fund_type='Venture Capital Fund' | 'Private Equity Fund' | 'Hedge Fund' | 'Other Investment Fund') 'Technology', 'Real Estate', 'Health Care', 'Energy', 'Financial Services', 'Manufacturing', 'Other'. Direct filing_id lookup is fastest (accession number). Substring filters on issuer_name, industry_group_type, investment_fund_type, and jurisdiction_of_inc enable topic-style queries. Combine federal_exemption + min_amount_sold for 'who's raising real money under 506(c)' analyses.

Returns safety recalls from federal agencies — drug recalls (FDA), medical device recalls (FDA), food/dietary supplement recalls (FDA), and consumer-product recalls (CPSC). Vehicle recalls (NHTSA) are NOT covered: source=nhtsa and the vehicle_make / vehicle_model filters return an INVALID_PARAMETER error rather than an empty list. Use this when the user asks about: recent recalls for a specific company or product, FDA Class I (most severe) recalls, food contamination recalls, drug shortages and recalls, or to add a 'product-safety event' flag to insider activity / 8-K filings / enforcement actions. Sources (filter via the `source` enum): fda_drug — openFDA /drug/enforcement.json. Drug recalls including prescription, OTC, biologics. Class I/II/III severity. fda_device — openFDA /device/enforcement.json. Medical device recalls (implants, diagnostics, equipment, software). Same classification scheme. fda_food — openFDA /food/enforcement.json. Food + dietary supplements. Pathogen contamination, allergen mislabeling, etc. cpsc — saferproducts.gov RestWebServices/Recall. Consumer-product recalls (clothing, electronics, toys, batteries, etc.). No severity classification; classification field is null. nhtsa — NOT COVERED: KeyVex does not serve NHTSA vehicle recalls yet; this source returns an INVALID_PARAMETER error. Cross-source pairing pattern: Recall → 8-K Item 7.01/8.01: pair with get_material_events Recall → insider sells: pair with get_insider_transactions Recall → SEC/DOJ follow-on: pair with get_enforcement_actions Recall → company filings: pair with get_proxy_filings (DEF 14A risk factors) Each record is one recall. Identifier format: `{source}-{recall_number}` (e.g., 'fda_drug-D-1234-2026'). FDA classifications: Class I — serious adverse health consequence or death Class II — temporary or reversible health consequence Class III — unlikely to cause adverse health consequence Source freshness (per-source publication cadence, not KeyVex bug): CPSC publishes within ~1-2 days; recent data flows hourly-fresh. openFDA's snapshot updates every ~10-14 days, and each snapshot carries recall_initiation_date values that LAG the snapshot date by another 30-45 days (the time between FDA classifying a recall and openFDA exposing it). Net: FDA records in this collection typically run ~4-6 weeks behind real-world recall dates, while CPSC is current. A default desc-by-date sort therefore looks CPSC-heavy at the top even when FDA matters more for the query. Filter by source='fda_*' to see FDA-only and avoid the skew. Searching recalling_firm / product_description: case-insensitive; 4+ characters match anywhere, 1-3 characters match a whole word ('3M', 'GE'). A term of 3+ characters searches EVERY recall; a 1-2 character term searches only the newest 5,000 recalls the query's other filters select (the oldest, with sort_order=asc), and coverage_warning says so. classification filter scope: 'classification' is an FDA-only field. CPSC records always have classification=null (CPSC doesn't use the FDA severity scheme). Filtering by classification excludes ALL CPSC rows by definition. The query response surfaces this with a notice in coverage_warning when the filter is set.

Returns Schedule 14A proxy filings — the document public companies send shareholders ahead of annual or special meetings. Each record is one filing carrying executive compensation tables, board nominations, shareholder proposals, auditor info, and voting matters. Use this when the user asks about: executive compensation, board elections, shareholder proposals, M&A votes, proxy contests, auditor changes, say-on-pay outcomes, or upcoming annual meetings. Coverage: the full DEF 14A family back to 2016 for the US public-company universe (sourced from EDGAR's complete quarterly full-index); a daily feed keeps it current. Rows are tagged with a company's PRIMARY common ticker — for dual-class issuers (e.g. GOOGL/GOOG, BRK-A/BRK-B) query by company_cik to retrieve every share class in one shot. Filing types (the four-form DEF 14A family): DEF 14A — Definitive proxy (the annual-meeting filing) DEFA14A — Additional materials (supplements to a prior DEF 14A) DEFM14A — Merger-related proxy (filed when shareholders vote on M&A) DEFR14A — Revised definitive proxy (amendments to a prior DEF 14A) Convenience flags derived from filing_type: is_merger_related — true for DEFM14A is_amendment — true for DEFR14A is_additional_materials — true for DEFA14A period_of_report population (IMPORTANT for filtering / sorting): - Recent-window rows (filed ~2024-onward via the daily feed / per-ticker pull): DEF 14A primaries ~100% populated (meeting/record date); DEFA14A/DEFM14A/DEFR14A typically EMPTY (correct-as-filed — SEC's submissions API leaves those reportDate fields blank). - Historical backfilled rows (the bulk of 2016-2024 depth, sourced from EDGAR's full-index): period_of_report is EMPTY for all form types — the index carries no report date. - Bottom line: filter/sort by filing_date for chronological queries; period_of_report is not reliably present across the collection. v1A is metadata-only: ticker, company name, CIK, filing type, dates, primary document URL. The proxy body is not extracted in v1. `primary_document_url` points agents at the source HTML for direct fetch when they need exec comp tables or proposal text.

Returns SEC Form 1-A filings — Regulation A+ 'mini-IPO' offering statements, 2015-06→present: companies raising up to $20M (Tier 1) or $75M (Tier 2) from the public without a full IPO. One record per filing with the issuer (SIC code, jurisdiction, year incorporated, employees, city/state), tier election, offering terms (security types, count, price, total aggregate amount, estimated net), service providers WITH FEES (underwriter, sales commissions, auditor, legal), and the issuer's summary financials from Part I (cash, assets, liabilities, equity, revenues, net income). Use this when the user asks about: Reg A / Reg A+ raises, mini-IPOs, small- cap capital formation, who's underwriting or auditing small offerings, or issuer financials before a raise. form family: '1-A' initial | '1-A/A' amendment | '1-A POS' post-qualification amendment (is_post_qualification) | -W withdrawals (is_withdrawal, metadata-level). One offering typically chains 1-A → 1-A/A… → qualification → 1-A POS updates — filter cik + sort asc to read it. Coverage (honest): the offering-statement family only. The Reg A+ periodic reports — 1-K annual (with actual proceeds raised), 1-SA semiannual, 1-Z exit — use different schemas and are a planned separate dataset. Offering circulars (253G) are prose documents — follow filing_index_url. Pure-publisher posture: issuer-reported Part I numbers as filed, parsed by KeyVex.

Returns SEC Form S-1 / S-3 / S-3ASR registration statements — securities offering registrations filed with the SEC. Use this when the user asks about: which companies are going public (IPO pipeline via S-1), shelf registrations (S-3 / S-3ASR — company registers securities to sell over multiple offerings without re-registering; large established issuers use the automatic S-3ASR variant), recent secondary offerings, registration amendments updating prior filings, or to bridge from a company name / ticker to the prospectus prose. Forms covered: S-1 — Initial registration (IPO + first-time registrants) S-1/A — Amendment to an S-1 S-3 — Shelf registration (issuers meeting reporting / market-cap criteria; lets them issue securities over time without re-registering each time) S-3/A — Amendment to an S-3 S-3ASR — Automatic shelf registration. The shelf form used by Well-Known Seasoned Issuers (large established companies like Apple, Ford, most of the S&P 500). Effective on filing. These issuers file S-3ASR, NOT plain S-3. Source: SEC EDGAR full-text search. Returns one record per filing, deduped by accession. Exhibit attachments (EX-10, opinion letters, fee tables, etc.) are filtered out — only the canonical form types are returned. v1A is metadata only. Each record has filer name + CIK + optional ticker + SEC file_number, state, SIC code(s), and URLs. Substantive prospectus content (offering size, share counts, use of proceeds, risk factors, financial statements) lives at primary_document_url — agents follow for the prose. SCOPE — covers S-1 (IPO), S-3 + S-3ASR (shelf, including WKSI auto shelves), plus /A amendments. S-8 employee-benefit-plan registrations, S-4 merger/acquisition registrations, and F-series foreign-issuer forms are NOT ingested. 424B prospectus supplements (offering takedowns off an existing shelf) are out of scope — query the shelf registration itself. Amendment chains: all amendments share the same sec_file_number as the original. Use sec_file_number filter to fetch an entire amendment chain. Pure-publisher posture: KeyVex doesn't derive 'likely-to-IPO' or 'price-target' signals from registration filings.

Returns congressional roll-call vote metadata (House + Senate) from api.congress.gov. Use this when the user asks about: recent votes in either chamber, votes on a specific bill, votes by date range, or to chain to per-member positions via the source_data_url. Sources: api.congress.gov v3 for House votes; senate.gov XML (legislative/LIS/roll_call_lists/) for Senate votes — joined into one collection. Captures roll-call (recorded) votes only — voice votes and unanimous-consent passages aren't roll calls and don't appear here. v1A returns vote-level metadata: chamber, roll call number, vote type, result, the legislation being voted on (linked via bill_id), and links to the Clerk's authoritative XML data. Per-member positions (yea/nay/present/not voting per bioguide_id) live in the XML at source_data_url; agents fetch that directly when they need member detail. v1.1 will add a separate roll_call_member_votes tool/ collection for queryable per-member positions. Vote identifiers are stable composite keys: '{chamber}-{congress}- {session}-{rcNumber}', e.g., 'house-119-1-240' or 'senate-119-1-15'. Common vote_type values: 'Yea-And-Nay' (regular recorded vote), '2/3 Yea-And-Nay' (suspension of rules, requires 2/3 majority), 'Recorded Vote', 'Quorum'. Common result values: 'Passed', 'Failed', 'Agreed to', 'Rejected', 'Motion Agreed To', 'Motion Failed'. When a vote is on a bill, legislation_type + legislation_number are populated and bill_id is set to the composite key — use that to join to get_bills. For procedural votes (motion to recommit, motion to adjourn, etc.), those fields may be empty. Amendment + Senate detail: House votes ON AN AMENDMENT carry amendment_number, amendment_author (sponsor + label), and amendment_type (e.g. 'HAMDT'). Senate votes carry vote_title (descriptive title — e.g. a confirmation or motion), measure (the specific measure a question references, e.g. 'S.Amdt. 5740'), and en_bloc_matters[] (one {issue, question, result} per matter when a batch of nominations is decided en bloc). All are empty / [] where not applicable.

Returns entries from the US Consolidated Screening List (CSL) — the combined feed of twelve federal export-screening lists. Use this when the user asks about: whether a company or person is on a US screening / sanctions / denied-party list (KeyVex republishes the lists; it is not a screening service), BIS Entity List members, Military End User designations, or to add a 'restricted party' flag to a federal contractor or foreign agent. The CSL unifies twelve lists (filter via source_short): SDN — Specially Designated Nationals (Treasury/OFAC) EL — Entity List (Commerce/BIS) DPL — Denied Persons List (Commerce/BIS) MEU — Military End User List (Commerce/BIS) UVL — Unverified List (Commerce/BIS) CMIC — Non-SDN Chinese Military-Industrial Complex Companies (Treasury) CAP — Capta List (Treasury) DTC — ITAR Debarred (State) ISN — Nonproliferation Sanctions (State) MBS — Non-SDN Menu-Based Sanctions List (Treasury) PLC — Palestinian Legislative Council List (Treasury) SSI — Sectoral Sanctions Identifications List (Treasury) Broader than get_ofac_sdn — the SDN list is just one source here. For the OFAC-SDN deep view use get_ofac_sdn; to search every US list at once use this tool. Cross-source pairing: pair with get_federal_contracts to flag a contractor that also appears on a screening list, and with get_foreign_agents for the foreign-entity overlay. Each record carries name + alt_names, the source list, sanctions programs, addresses, distinct countries, and identification documents.

Returns SEC comment-letter correspondence: form UPLOAD (the SEC's letter TO the company — the questions) and CORRESP (the company's response). The Division of Corporation Finance sends these during filing reviews; they're released ~20+ business days after the review closes. Coverage 2005→present. Use this when the user asks about: whether a company is (or was) under SEC review, accounting-quality red flags before they become enforcement, the back-and-forth around an IPO registration, or to pair with fundamentals / insider activity ('were insiders selling while the SEC was asking questions?'). Reading a thread: filter by ticker or cik, sort date_filed asc — a review is an alternating UPLOAD/CORRESP chain; the final short UPLOAD is typically the 'review complete' letter. v1A is metadata-only: follow filing_index_url for the letter text. released_date is set on records captured from EDGAR's daily indexes (the dissemination day); older backfilled records carry only date_filed (the letter's own date) — dissemination day isn't recoverable historically and KeyVex never fabricates it. A comment letter is ROUTINE, not an accusation — most large filers get reviewed on a cycle (Sarbanes-Oxley §408 requires review at least every 3 years). Signal comes from thread LENGTH, topic, and recency, which agents judge from the letter text. Pure-publisher posture: EDGAR index records as published.

Returns SEC Fails-to-Deliver (FTD) rows — daily settlement failures by ticker / CUSIP / date. Each row is one ticker on one settlement date where a clearing-member's short sale FAILED to deliver shares. Signal value: persistent FTDs are a contrarian short-squeeze leading indicator. When the daily FTD quantity spikes on a ticker, it often means naked short pressure overwhelming locate supply or settlement / locate mechanism breaking down. The Reg SHO Threshold Securities list (FTDs > 0.5% of issued shares for 5+ consecutive days) is a derived view; this tool exposes the underlying daily data. Source: SEC bi-monthly cnsfails<YYYYMM><a|b>.zip files at sec.gov/files/data/fails-deliver-data/. Published ~1 week after each half-month settlement period. Coverage: every U.S.-listed security with a recorded settlement failure during the period. Killer query patterns: - Daily FTD history for a ticker: ticker='GME' + sort_by='settlement_date' - Largest FTDs this month: min_value=1000000 + sort_by='fail_value' - Squeeze setup candidates: min_quantity=100000 + recent dates - Look-up by CUSIP: cusip='B6S7WD106' (foreign issuers, complex names) Derived field: fail_value = quantity_fails × price (dollar magnitude of the failure on that day). Reference price comes from the SEC's posted value at settlement. Note: FTDs are bi-monthly batch-published, not real-time. SEC releases each half-month batch (cnsfails<YYYYMM>a = days 1-15, b = 16-end) roughly 2-4 weeks AFTER that half-month period closes, so the most recent settlement date can be 2-4 weeks behind today (e.g. in mid-June the latest published batch is first-half-May, with settlement dates through ~May 15). That apparent lag is the SEC publish cadence, not a KeyVex freshness gap.

Returns SEC Schedule TO filings — public tender offer disclosures. Use this when the user asks about: who's bidding to acquire company X, what M&A offers are in flight, share buyback announcements, amendments to existing tender offers (price increases / extensions), or to pair with 13D activist stakes for the 'stake → bid' story. Source: SEC EDGAR full-text search. Forms covered: SC TO-T (third- party tender offer — someone outside the company bidding for shares), SC TO-T/A (amendments), SC TO-I (issuer tender offer — company buying back its own shares), SC TO-I/A (issuer amendments). v1 returns filing metadata only — bidder + target + form type + filing date + URL. Offer price, shares sought, and expiration date live inside the HTML attachment at primary_document_url; agents follow that URL to read the substantive terms. Amendment filings share the same target/bidder/file_number as the original offer; use file_number to group an amendment chain. Pure-publisher posture: KeyVex does not derive 'likely to close' or 'expected premium' signals. The data here is what was filed, no more.

Returns Treasury security auctions — Bills (≤1yr), Notes (2-10yr), Bonds (20-30yr), TIPS (inflation-protected), and FRNs (floating-rate). Each record is one CUSIP issuance with announcement metadata + post- auction results. Key signal fields agents care about: - bid_to_cover_ratio: demand. >2.5 strong, <2.0 weak. - high_yield / average_yield: market clearing rate. - direct_bidder / indirect_bidder breakdowns: domestic vs foreign demand. - soma_holdings + soma_included: Fed System Open Market Account allocation. A live measure of Fed QE/QT activity on each issue. Records have a two-stage lifecycle: announcement (results fields null) → post-auction (full results populated). Idempotent saves on cusip + auction_date overwrite cleanly when results publish. Security types: 'Bill', 'Note', 'Bond', 'TIPS', 'FRN', 'CMB' (cash- management bill). Use security_type filter to focus on one term group. Note: Treasury reports TIPS and FRNs under security_type Note/Bond with an inflation-indexed / floating-rate flag (not as their own type); filtering security_type:'TIPS' or 'FRN' here resolves to those flags for convenience.

Identifier-driven cross-collection fan-out search. Pass one or more entity identifiers — ticker, bioguide_id, company_cik, recipient_uei, company_name, or cusip — and this tool queries every collection where that field is indexed, returning results grouped by source in one envelope. Congressional-trade and annual financial disclosure results come from STOCK Act and Form 278 filings. The same filings are published free as news at https://keyvex.com/disclosures under 5 U.S.C. § 13107(c). Use this for high-level 'tell me everything about X' questions before drilling into specific source tools. Replaces 6-10 sequential tool calls with a single fan-out. Identifier coverage: - ticker → 13 collections (company_profile, insider_trades, institutional_holdings, congressional_trades, planned_insider_sales, initial_ownership_baselines, activist_ownership, material_events, proxy_filings, xbrl_fundamentals, tender_offers, registration_statements, nport_holdings) - bioguide_id → 2 collections (congressional_trades, annual_financial_disclosures). In both, a record's party is the party the member held on the date of the record (a trade's transaction date, a disclosure's filing date); for a date outside the member's terms in office, the party of their nearest term (the last one before that date, or the first one after it). - company_cik → 11 collections (company_profile, insider_trades, planned_insider_sales, initial_ownership_baselines, activist_ownership, material_events, proxy_filings, xbrl_fundamentals, private_placements, registration_statements, nport_filings) - recipient_uei → 1 collection (federal_contracts) - company_name → 5 name-keyed collections (federal_contracts, enforcement_actions, consumer_complaints, product_recalls, fda_approvals) AND auto- resolves to ticker + company_cik via EDGAR's catalog, cascading into every ticker/CIK adapter above. The unlock for 'tell me everything about Wells Fargo' hitting up to 20 collections in one call. (lobbying_filings is excluded from the fan-out — its 51K-record substring scan is too slow for parallel federation; call get_lobbying_filings directly instead.) enforcement_actions here returns ENFORCEMENT only, as get_enforcement_actions does by default; call get_enforcement_actions with category="all" for the agencies' general news too. When a regulator's site refuses KeyVex (CFTC since 2026-10-07), the freshness block says which records are missing. - cusip → 4 collections (institutional_holdings, activist_ownership, nport_holdings, treasury_auctions) Multiple identifiers can be combined to narrow each source's query (e.g., ticker + bioguide_id will filter congressional_trades by both). Name resolution: when only company_name is supplied, the resolved ticker and CIK come from EDGAR's company_tickers_exchange.json (US-listed names only). Foreign-only or private companies won't resolve to a ticker, but name-keyed collections (CFPB, lobbying, etc.) still receive the substring filter directly. Performance note: company_name fan-out includes substring-filtered collections (lobbying_filings 51K+, federal_contracts) which scan a 5K- record window per source. The full cascade can take 10-40s depending on Firestore region. When latency matters more than coverage, pass `sources` to whitelist only the ticker/CIK-keyed adapters (typically <1s total). Per-source result count is capped via per_source_limit (default 5, max 50). One slow or failing source returns an error block instead of blocking the rest — check sources_queried vs sources_with_results to see what landed. When you need full result counts and richer filters on one specific collection, call that collection's dedicated tool directly afterward.