SEC Filings Scraper: 10-K, 8-K, Full-Text Search

SEC Filings Scraper: 10-K, 8-K, Full-Text Search

SEC filings scraper for 10-K, 10-Q, 8-K and every other EDGAR form: one row per filing, by ticker, CIK, company name, accession number, full-text search or date window, with no login, no API key and no proxy setting.

Each Filing row carries the form and what it is, the filing date and the acceptance time, the report period (with fiscal year and quarter when the filer's fiscal year end is known), the 8-K item codes with their names, the filer's industry (SIC), state of incorporation, tickers and exchanges, file and film number, size, the links to the index page, the folder, the full submission, the main document and the inline XBRL viewer, and a one sentence summary built from those fields. Two paid add-ons read further: the main document as plain text with its report sections, and the XBRL figures a 10-K, 10-Q, 20-F or 40-F reported for its own period. Use it as an SEC filings API for your own code (curl, Node and Python examples below), as a scheduled feed of new 8-Ks, or as a CSV export for a spreadsheet.

Why this one

• Pay per filing delivered, plus a small start fee per run. One Filing event per filing row. Company profile rows, filing details, status rows and the run summary are free. An unknown ticker, CIK or name, a filing that is not found, a filing your filters leave out and a filing found twice cost nothing, and every miss is named in a free row.
• Full-text search past the 10,000 result cap. The SEC's full-text search lists at most 10,000 results per query. Newest first, this Actor splits a search by date and, for one large day, by form, and reads every part. Measured on 2026-11-01, the parts of one day with more than 10,000 results added up to the day's total (10,319 results).
• Complete company histories. Each company's whole filer record is read, its older pages included, back to 1994 for the oldest filers. A date window reads only the pages it reaches.
• Every filing of a date window. Leave the targets empty and you get every filing of the last 7 days (or your window) of the forms you choose. Ready-made searches cover going concern, material weakness, restatement, cybersecurity incident, layoffs, guidance withdrawn, CEO departure, chapter 11, share buyback, activist investor, artificial intelligence and tariffs.
• Screen with several searches at once. Give several queries or presets: each filing is one row that lists every one that found it (matchedQueries, matchedPresets), charged once, and every row carries a materiality, novelty and signal score from published weights with the tokens that drove them. The run also stores its filings ranked by signal score.
• 8-K items with names, 16 event categories and risk flags. Earnings, leadership change, cybersecurity incident, auditor change and the rest are filters and fields of the row, read from the SEC's own item codes and form types.
• Fast and polite. On the platform at 256 MB, the newest 1,000 filings of the window 2026-10-22 to 2026-11-01 took 15 seconds, and 989 filings of 20 large companies since 2024 with their 20 company rows took 23 seconds (2026-11-01 UTC). The SEC is asked at most 5 times per second, half its published limit of 10, with a declared User-Agent.

Quick start

1. For one company, put its ticker into Tickers (the prefill is AAPL with the form types 10-K, 10-Q and 8-K). For a filer without a ticker, use CIK numbers.
2. Or put words or "exact phrases" into Full-text search queries, or pick Ready-made searches.
3. Or leave every target empty for every filing of the last 7 days; set Form types, Filed from and Filed until (YYYY-MM-DD) or Days back to change what is read.
4. Optional: narrow with 8-K item codes or 8-K event categories, and turn on Filing details (free), Document text or Financial data (paid add-ons).
5. Set Maximum filings (the prefill is 10; 100 is used when the field is left out) and click Start. Rows appear in the Dataset tab; export JSON, CSV or Excel, or read them through the API.

What you get

Every row has a type: filing (the paid row), company (a free profile of each company read by ticker, CIK or name), status (free, one per kind of miss or input problem) and summary (free, the last row of every run).

Filing row

One row per filing, 62 fields. The blocks:

• Identity: accessionNumber, form, formDescription, rootForm, isAmendment, filedAt, acceptedAt, reportDate, fiscalYear, fiscalPeriod, act, fileNumber, filmNumber, sizeBytes.
• Filer: cik, companyName, tickers, exchanges, filers (cik, name, role), sic, sicDescription, stateOfIncorporation, fiscalYearEnd, entityType, businessLocation.
• Events: items, itemDescriptions, eventCategories, riskFlags.
• Documents and links: isXBRL, isInlineXBRL, primaryDocument, primaryDocDescription, filingIndexUrl, filingFolderUrl, fullSubmissionUrl, primaryDocumentUrl, inlineXbrlViewerUrl, companyFilingsUrl.
• Search: searchQuery (the first query that found the filing), matchedQueries (every query and preset of the run that found it), matchedPresets, matchedDocuments (fileName, type, description, score) and relevanceScore, on rows found by a full-text search.
• Signals: materialityScore, materialityLevel, noveltyScore, noveltyLevel, noveltyBasis, priorSimilarInRun, signalScore, signalDrivers, alertSeverity, on every filing row (see Signals below).
• Add-ons: filingDetails, documentText, financialData, addOnNotes (why an add-on was left off a row).
• Run: summary, source (filer_record, full_text_search or index_page), subjectKey, scrapedAt, type.

riskFlags names what the SEC's codes say: bankruptcy, cybersecurity_incident, material_impairment, delisting_notice, auditor_change, non_reliance_restatement, change_in_control, officer_or_director_change from 8-K items, and late_annual_report, late_quarterly_report, deregistration, delisting from the form type.

Fill rates

Measured on two runs of this Actor on the platform (2026-11-01 UTC). Company mode rows (989): the filings since 2024-01-01 of 20 large companies, up to 50 each (by form with its amendments: 480 Form 4, 114 Form 144, 87 8-K, 52 424B2, 33 10-Q and 223 of 32 other forms). Date window rows (1,000): the newest 1,000 filings of the window 2026-10-22 to 2026-11-01, without filing details (335 Schedule 13G, 261 Form 4, 168 8-K, 103 6-K, 91 10-Q and 42 of 5 other forms). A field counts as filled when it is not null, not an empty text and not an empty list. The rates follow the mix of forms: a Form 4 has no report period and no 8-K items.

Field	What it holds	Company mode rows (989)	Date window rows (1,000)
form, filedAt, companyName, primaryDocument, primaryDocumentUrl	Form, filing date, filer, main document	100%	100%
acceptedAt	Acceptance time (UTC)	100%	0% (100% with filing details)
sizeBytes	Size of the whole submission	100%	0% (100% with filing details)
isInlineXBRL	Inline XBRL flag	100%	0% (100% with filing details)
isXBRL	XBRL flag	100%	0%
reportDate	Report period, all rows	66.9%	64.4%
reportDate	Report period, 10-K and 10-Q rows	100%	100%
fiscalPeriod	FY, Q1 to Q3, 10-K and 10-Q rows	100%	0%
items	8-K item codes, 8-K rows	100%	100%
tickers	Tickers of the filer	100%	70.4%
sic	Industry code	100%	39.1%
stateOfIncorporation	State of incorporation	100%	66.8%
fiscalYearEnd	Fiscal year end (MMDD)	100%	0%
businessLocation	City and state of the filer's business address	100%	39%
act	Securities act	49%	0%
fileNumber, filmNumber	SEC file and film number	49.2%, 48.9%	39.7%, 39.7%
primaryDocDescription	Description of the main document	81.8%	58.6%

The "with filing details" figures are from a run of 30 8-K filings of a 5 day window with includeFilingDetails on (2026-11-01 UTC). The filers of a Form 4 or a Schedule 13G are often people, who have no industry code, no tickers and no fiscal year end. Since 2026-11-06 the business location of every filer is delivered as the SEC serves it (the figures above were measured before, when it was left out for a filer without an industry code); for many people the SEC serves none.

Example Filing row (real run 5pASsU42HyuViOnSf of 2026-11-02, the 20 company bulk run; shortened: 42 of its 62 fields):

{
  "type": "filing",
  "accessionNumber": "0000320193-26-000018",
  "form": "8-K",
  "formDescription": "Current report",
  "isAmendment": false,
  "filedAt": "2026-07-30",
  "acceptedAt": "2026-07-30T20:30:28.000Z",
  "reportDate": "2026-07-30",
  "cik": "0000320193",
  "companyName": "Apple Inc.",
  "tickers": ["AAPL"],
  "exchanges": ["Nasdaq"],
  "sic": "3571",
  "sicDescription": "Electronic Computers",
  "stateOfIncorporation": "CA",
  "fiscalYearEnd": "0926",
  "businessLocation": "CUPERTINO, CA",
  "items": ["2.02", "9.01"],
  "itemDescriptions": ["Results of Operations and Financial Condition", "Financial Statements and Exhibits"],
  "eventCategories": ["earnings"],
  "riskFlags": [],
  "materialityScore": 35,
  "materialityLevel": "medium",
  "noveltyScore": 99,
  "noveltyLevel": "high",
  "noveltyBasis": 34,
  "priorSimilarInRun": 0,
  "signalScore": 61,
  "signalDrivers": ["high_novelty", "first_of_its_kind_in_run"],
  "alertSeverity": "medium",
  "act": "34",
  "fileNumber": "001-36743",
  "filmNumber": "261223821",
  "sizeBytes": 417360,
  "isXBRL": true,
  "isInlineXBRL": true,
  "primaryDocument": "aapl-20260730.htm",
  "filingIndexUrl": "https://www.sec.gov/Archives/edgar/data/320193/000032019326000018/0000320193-26-000018-index.htm",
  "primaryDocumentUrl": "https://www.sec.gov/Archives/edgar/data/320193/000032019326000018/aapl-20260730.htm",
  "summary": "Apple Inc. filed an 8-K (current report) on 2026-07-30 for the period ending 2026-07-30. Items: 2.02 Results of Operations and Financial Condition; 9.01 Financial Statements and Exhibits.",
  "source": "filer_record",
  "scrapedAt": "2026-11-02T03:37:10.295Z"
}

Signals

Every filing row is scored by fixed rules, the same on every run, with nothing read from the filing's text:

• materialityScore (0 to 100): the form's weight plus the weight of every 8-K item it states plus the weight of every preset it matched, at most 100. Forms (with their amendments): 10-K, 20-F and 40-F 30; 10-Q and 8-K 20; 6-K 15; S-1, F-1, S-4 and F-4 25; 424B4 15; Schedule 13D 35, Schedule 13G 15; tender offers (SC TO-T, SC 14D9) 35; proxy contest DEFC14A 40, PREC14A 35; DEF 14A 15, PRE 14A 10; NT 10-K and NT 20-F 40, NT 10-Q 35; 15-12B, 15-12G, 25 and 25-NSE 40, 15-15D 35; Form 4 10; Forms 3, 5, 144 and D 5; any other form 10. 8-K items: 1.03 bankruptcy 60, 4.02 non reliance 55, 1.05 cybersecurity 50, 3.01 delisting 45, 5.01 change in control 40, 2.06 impairment 35, 4.01 auditor change 35, 2.01 and 2.05 30, 1.04 30, 2.04 and 5.02 25, 1.01 and 1.02 20, 5.06 20, 2.02 and 2.03 15, 3.03 15, 3.02, 5.03, 5.04 and 5.05 10, 5.07, 5.08, 7.01 and 8.01 5. Presets: bankruptcy 40, going concern 35, material weakness 30, restatement 30, cybersecurity incident 30, guidance withdrawn, CEO departure and activist investor 20, layoffs 15, share buyback and tariffs 5. materialityLevel: critical from 75, high from 50, medium from 30, else low.
• noveltyScore (0 to 100): how rare the filing is among the filings this run delivered before it: 40 x (1 - the share of those with its form) + 40 x (1 - the share of those with its event kind) + 20 when the same company has no filing of that kind earlier in the run. The event kind is the 8-K event category (the rarest one), else the form. noveltyBasis says how many filings it was compared with; the first row of a run has none, so its novelty is null. noveltyLevel: high from 75, medium from 45, else low. priorSimilarInRun counts the same company's earlier filings of that kind in the run.
• signalScore: 0.6 x materiality + 0.4 x novelty, rounded (materiality alone on the first row). signalDrivers lists why: the row's riskFlags, presets weighted 20 or more, high_materiality, high_novelty, first_of_its_kind_in_run. alertSeverity is the materiality level (critical, high, medium), low for a low row with a risk flag, else none.

The dataset keeps the order the filings were read (newest first). The 1,000 highest signal scores of the run, highest first, are stored in the run's key-value store as SIGNAL_RANKING (rank, signalScore, materialityLevel, noveltyLevel, filedAt, form, companyName, cik, accessionNumber, filingIndexUrl), linked from the run's output and named in the summary row.

Company row (free)

With includeCompanyRows on (the default), every company read by ticker, CIK or name gets one free company row: cik, name, entityType, tickers, exchanges, sic, sicDescription, industryOffice, category (for example Large accelerated filer), ein, stateOfIncorporation, stateOfIncorporationDescription, fiscalYearEnd, phone, website, investorWebsite, businessAddress, mailingAddress, formerNames (with from and to dates), isInsiderFiler, hasInsiderFilingsAsIssuer, personalDataWithheld, filingsOnRecord, olderFilingPages, companyFilingsUrl, subjectKey. For Apple in the run above: filingsOnRecord 2271, phone (408) 996-1010, three former names back to APPLE COMPUTER INC (1994-01-26). Every filer's addresses, phone and EIN are delivered as the SEC serves them, a filer without an industry code (who may be a person) included; personalDataWithheld stays in the row and is always null.

Four ways to ask

• Companies (tickers, ciks, companyNames): each company's filer record, newest first, with its older pages as far as the date window reaches. A ticker is looked up in the SEC's list of listed tickers; a name is used only when exactly one listed company carries it. For a list of two or more companies with a date window from 2001 on, each batch of 100 companies is first checked with one page of the full-text index; when that page holds the batch's whole answer (100 filings or fewer in the window), only the companies it lists are read, otherwise every company of the batch is read.
• Filings (accessionNumbers): each filing's own index page, by accession number or EDGAR link, delivered as asked (date, form, item and size filters cannot be combined with them).
• Full-text search (searchQueries, searchPresets): the SEC's full-text search with its own syntax ("exact phrase", OR, a leading minus to exclude), optionally within companies given by tickers or ciks, newest first or by relevance. One row per filing, with the documents that matched; a filing that several queries or presets find is one row that lists them all in matchedQueries.
• Date window (no target): every filing of the window, newest first, from the full-text index (one search per part, no query).

Add-ons

• Filing details (free, includeFilingDetails). Reads each filing's index page into filingDetails: every document and exhibit (sequence, description, name, type, sizeBytes, url, group), every filer with its role (filer, issuer, subject, reporting, filed_by) and its sic, ein, stateOfIncorporation, fiscalYearEnd, fileNumber, filmNumber, act and addresses, and totalSizeBytes. It also fills acceptedAt, sizeBytes and isInlineXBRL on search and window rows. One extra request per filing makes a run slower.
• Document text (paid, includeDocumentText). Reads the main document into documentText: text (up to maxTextCharacters, 200,000 by default), characters, totalCharacters, truncated, and sections (item, title, characters, text) for 10-K, 10-Q, 20-F and 40-F reports (Item 1A Risk Factors, Item 7 MD&A and the rest) and the items of an 8-K or 6-K. For an 8-K or 6-K it adds the press release (exhibit 99.1) in pressRelease, the sentences that state an outlook with a number and a period in guidanceSentences, and a count of finance words in sentiment (words, positive, negative, uncertain, net), from fixed word lists. For a Schedule 13D or 13G filed as XML it reads the stake of each reporting person into ownership (reportingPersons with aggregateAmountOwned, percentOfClass, sole and shared voting and dispositive power). With textChunkSize the text is also cut into chunks at paragraph or sentence ends, for retrieval and AI pipelines.
• Financial data (paid, includeFinancialData). Attaches to a 10-K, 10-Q, 20-F or 40-F the XBRL figures it reported for its own period, in financialData (taxonomy, concept, label, value, unit, periodStart, periodEnd, fiscalYear, fiscalPeriod, frame): revenue, net income, operating income, earnings per share, assets, liabilities, equity, cash, operating cash flow and long-term debt (US GAAP; the IFRS set for a 20-F or 40-F without US GAAP figures), plus shares outstanding from the cover, or the concepts you list in financialConcepts. A 10-Q gives its quarter and its year to date.

A real Financial data entry (Apple's 10-Q for the quarter ended 2026-06-27, platform run of 2026-11-01):

{"taxonomy": "us-gaap", "concept": "RevenueFromContractWithCustomerExcludingAssessedTax", "label": "Revenue from Contract with Customer, Excluding Assessed Tax", "value": 109417000000, "unit": "USD", "periodStart": "2026-03-29", "periodEnd": "2026-06-27", "fiscalYear": 2026, "fiscalPeriod": "Q3", "frame": "CY2026Q2"}

An add-on that cannot deliver for a row (a form without XBRL figures, a document that cannot be read, no room left under your charge limit) is left off that row with a note in addOnNotes and is not charged.

Status rows (free)

Every input problem and every miss gets a free type: "status" row with scope, status and a message in plain words. Misses of one kind share one row with the items and their count.

status	scope	What it means
invalid_input	input	A setting that cannot be used (a date, a filter value, a cap, an unknown field name, own proxy URLs). The run stops before any paid work; only the run's start fee is charged.
unusable_item	input	Items of a target list that are not a ticker, CIK, name, accession number or query; their positions are listed and the other items run.
not_searchable	input	Form types with a slash other than the amendment mark, which the full-text search cannot filter on.
unknown_ticker	miss	Tickers that are not in the SEC's list of listed tickers.
unknown_cik	miss	CIKs the SEC does not know.
name_not_found	miss	Names that match no listed company name.
name_ambiguous_or_inexact	name	Names no listed company carries exactly; names lists each one (input, position) with up to 10 near candidates (cik, name, ticker, exchange). Nothing is guessed.
same_company_twice	miss	A company asked twice (for example by ticker and by CIK); its filings are delivered once.
accession_not_found	miss	Filings not found under the CIK of their link.
accession_not_found_or_agent_filed	miss	Bare accession numbers not found under the CIK their first ten digits name; for a filing an agent filed, give the EDGAR link.

A real status row (platform run of 2026-11-02; shortened: key and scrapedAt left out, two of the candidates shown):

{"type": "status", "scope": "name", "status": "name_ambiguous_or_inexact", "message": "1 name(s): no listed company carries exactly this name. Run again with the CIK of the company you mean (candidates below). Nothing was charged for them.", "names": [{"input": "Apple", "position": 1, "candidates": [{"cik": "0000320193", "name": "Apple Inc.", "ticker": "AAPL", "exchange": "Nasdaq"}, {"cik": "0001418121", "name": "Apple Hospitality REIT, Inc.", "ticker": "APLE", "exchange": "NYSE"}]}], "count": 1}

The last row of every run is a free type: "summary" row: status (complete, partial or invalid_input), mode, stoppedBy (for example max_items, charge_limit, work_allowance, source_unavailable), filingsDelivered, window, chargedEvents, counts, coverage (per company, search or window: status, delivered, oldestFilingDateReached, reason), sourceFailures, requests, allowance, signalRanking (the ranking record, its entries and whether it was saved), resumed and message. status is complete only when nothing stopped the run and every company, search and page it had to read was read.

How much does it cost to scrape SEC filings?

You pay per Filing row delivered plus a small start fee per run. The two paid add-ons are charged per row, only when they deliver. Company profile rows, filing details, status rows and the run summary are free.

Event	FREE	BRONZE	SILVER	GOLD
Filing (one SEC filing delivered as a row)	$0.11 / 1,000	$0.105 / 1,000	$0.10 / 1,000	$0.095 / 1,000
Document text (add-on: the main document read into text)	$0.70 / 1,000	$0.65 / 1,000	$0.58 / 1,000	$0.50 / 1,000
Financial data (add-on: XBRL figures of the report's own period)	$0.70 / 1,000	$0.65 / 1,000	$0.58 / 1,000	$0.50 / 1,000
Actor start (once per run)	$0.002	$0.002	$0.002	$0.002

Platinum and Diamond plans pay the Gold price. The Actor start event is a flat fee with no tier discount, charged once per run whatever the run delivers (it is charged per GB of memory, and this Actor runs at 256 MB).

The Store pricing card shows these same prices per 1,000 events: "$0.11 / 1,000" on the Filing row means one filing costs 0.011 cents, and "$0.70 / 1,000" on the Document text row means one document costs 0.07 cents.

Worked examples (start fee included):

Run	FREE	BRONZE	SILVER	GOLD
A search that finds nothing	$0.002	$0.002	$0.002	$0.002
10 filings (the prefill)	$0.0031	$0.00305	$0.003	$0.00295
1,000 filings	$0.112	$0.107	$0.102	$0.097
10,000 filings	$1.102	$1.052	$1.002	$0.952
100 10-K and 10-Q filings, each with document text and financial data delivered	$0.153	$0.1425	$0.128	$0.1115

What is free (no event other than the start fee): an invalid input, an unknown ticker, CIK or name, an ambiguous name, a filing that is not found, a filing your filters leave out, a filing found twice (by two queries, or a company asked twice), a filing past maxFilingsPerForm, every company row, filing details, every status row and the run summary. A document text or financial data read that delivers nothing is not charged. Rows are charged only after they are in your dataset, and your maximum charge per run is respected: a filing that does not fit is not delivered, and the summary says stoppedBy: charge_limit. A platform restart never charges a filing twice (the platform itself charges the start fee once more when a run is resurrected).

Input

Field	Default	What it does
tickers	empty (prefill AAPL)	Stock tickers such as AAPL, MSFT or BRK-B (BRK.B is read as BRK-B), matched through the SEC's list of listed tickers.
ciks	empty	SEC Central Index Keys of companies, funds or people, with or without leading zeros, or an EDGAR company link. Use a CIK for any filer without a ticker.
companyNames	empty	Names of listed companies, for example Apple Inc.; used only when exactly one listed company carries the name (case, punctuation and spacing aside).
accessionNumbers	empty	Filings by accession number (0000320193-26-000011) or EDGAR filing link, read as asked. Cannot be combined with companies or searches.
searchQueries	empty	Full-text queries in the SEC's syntax: "exact phrase", OR between alternatives, a leading minus to exclude a word. Narrow them with tickers or ciks, form types and dates.
searchPresets	none	Ready-made searches: going_concern, material_weakness, restatement, cybersecurity_incident, layoffs, guidance_withdrawn, ceo_departure, bankruptcy, share_buyback, activist_investor, artificial_intelligence, tariffs.
formTypes	every form (prefill 10-K, 10-Q, 8-K)	EDGAR form types to keep, for example 10-K, 8-K, 4, S-1, DEF 14A, SC 13D. A form includes its amendments unless includeAmendments is false; 10-K/A alone means amendments only.
includeAmendments	true	Keep amended filings (form types ending in /A).
dateFrom, dateTo	empty	First and last filing date, YYYY-MM-DD. Without dateFrom: a company's whole history, a search from 2001 on, the last 7 days for a date window.
daysBack	empty	Instead of dateFrom: the last N days on the SEC's calendar (US Eastern); 0 is today only.
items8K	none	Keep 8-K filings that report at least one of these items, for example 2.02, 5.02, 1.01.
eventCategories	none	Groups of 8-K items: earnings, material_agreement, mergers_acquisitions, bankruptcy, cybersecurity, debt_financing, restructuring, delisting, equity_sale, auditor_change, restatement, leadership_change, governance, shareholder_vote, regulation_fd, other_events.
minFilingSizeKb	none	Keep filings whose whole submission is at least this many KB (searches and windows read each index page first).
companyNameContains	empty	Keep search and window rows where a filer's name contains this text.
sortBy	newest	newest (every match, split past 10,000) or relevance (the search's own score, at most 10,000 results).
includeFilingDetails	false	Free: documents, filer roles and the acceptance time from each filing's index page.
includeDocumentText	false	Paid add-on: the main document as text, sections, press release, 13D or 13G stake.
maxTextCharacters	200,000	Characters of text kept per document, 1,000 to 2,000,000; the full length is reported.
textChunkSize	0	Chunks of about this many characters (500 to 100,000) for retrieval; 0 means none.
includeFinancialData	false	Paid add-on: XBRL figures of a 10-K, 10-Q, 20-F or 40-F for its own period.
financialConcepts	default set	Up to 40 XBRL concepts instead of the default set, for example Revenues, us-gaap:NetIncomeLoss, ifrs-full:Revenue.
includeCompanyRows	true	One free profile row per company read.
includePersonalData	false	All public data is always delivered, so this setting has no effect (it is kept so saved inputs keep working).
maxItems	100 (prefill 10)	The most Filing rows this run delivers and charges, 1 to 100,000.
maxFilingsPerCompany	none	The most filings per company, so a list of companies shares the run.
maxFilingsPerForm	none	The most filings per form type (a form and its amendments together); the rest are skipped free.
maxRequestsPerSecond	5	How fast the run asks the SEC, 1 to 5.
proxyConfiguration	off	Not needed. With Apify Proxy on, every request goes through one Apify datacenter proxy session from the start. Own proxy URLs are not used.

Which targets are present decides the kind of run: when any target field is present, even empty, the run reads those targets and is never widened into a date window (an empty list is a free invalid_input row). An input field this Actor does not know stops the run with a free row, because it could be a filter under another name.

Limits

• Date window coverage. The date window and the full-text search read the SEC's full-text index. On a day we compared with the SEC's daily index (2026-10-29) it held every 8-K, 10-Q, 10-K and Form 4; about 1.2 percent of the day's filings were not in it: SEC staff letters (UPLOAD, CORRESP), notices of effectiveness (EFFECT) and draft registrations (DRS). Company mode reads each company's complete filer record and gets those too, except in one case: when a batch of 100 listed companies filed 100 filings or fewer in a date window from 2001 on, only the companies the full-text index lists are read, so a company whose only filings in the window are of those kinds gets no rows.
• How far back. Filer records reach back to 1994 for the oldest filers; the full-text search and the date window cover filings from 2001 on.
• Several queries, newest first, are read side by side one filing day at a time: each day is read to its end in every search before its rows are delivered, so a run with several broad queries reads a whole day of each before its first row of that day. A run cut by a cap (maxItems, the charge limit, the free work room used up) or by one search failing delivers the filings already read, as far as the cap allows, and their rows can list fewer queries than a complete day would; the summary marks those searches partial. After a stop (an abort, a timeout, the SEC refusing the run) nothing of the cut day is delivered or charged. In relevance order the queries are read one after another and a row names the query that delivered it. With several searches and a charge limit, the free work room a run may use before its rows are charged is counted as if every filing took the add-ons you asked for at their full price, so near the limit such a run can stop earlier with stoppedBy: work_allowance (the filings read so far are delivered; nothing is charged past what it earns).
• Relevance order reads at most 10,000 results and says so in the summary. Newest first, a single form on a single day with 10,000 or more matches ends that part with a note naming it.
• Acceptance time. Rows read by ticker, CIK, name or accession number carry it; search and window rows carry it with filing details. isXBRL is filled only on rows read by ticker, CIK or name, and fiscalYear and fiscalPeriod only on 10-K, 10-Q, 20-F and 40-F rows whose filer's fiscal year end is known (rows read by ticker, CIK, name or accession number).
• Company names are matched among listed companies only (the SEC's list of tickers). For any other filer, give its CIK.
• Bare accession numbers are looked up under the CIK their first ten digits name: that always works when the company filed for itself, and often (not always) for a filing an agent filed; when it does not, the free row asks for the EDGAR link, which always works.
• Documents larger than 12 MB are not read, and the text of one row is kept under about 3,000,000 characters in total.
• Per run: up to 500 tickers, CIKs, names and accession numbers together, 20 queries of up to 500 characters, 200 values per filter list.
• Free work is limited. Reads that end without a Filing row earn nothing, so a run does a bounded amount of them, and every filing it delivers earns it more room. When the room is used up the run stops starting new reads, delivers what it has and ends with stoppedBy: work_allowance. The add-ons have their own room: when it is spent, rows go without the add-on, with a note, uncharged.
• Pace: at most 5 requests per second. Requests go out from the run's own address; if the SEC refuses it three times in a row, the rest of the run goes through one Apify datacenter proxy session at the same pace, and the summary says so. If the SEC refuses that too, the run stops with stoppedBy: source_unavailable and nothing more is charged.

What this Actor does NOT do

• No "new only" mode and no memory between runs: each run reads what its input asks. For a feed, schedule it with daysBack or dateFrom; a filing inside two runs' windows is delivered by both.
• No Form 4 transaction rows, no 13F holding rows and no XBRL statement rows: a Form 4 or a 13F-HR comes as one Filing row, and the Financial data add-on attaches only the figures a report filed for its own period.
• No AI summaries: summary is one sentence built from the row's own fields, and the outlook sentences and word counts come from fixed word lists.
• No earnings call transcripts: they are not filed with the SEC.
• No scores from the filing's text and no company memory: the signals come from the form, the 8-K item codes, the presets a filing matched and the filings of the same run. Nothing is remembered between runs (rule above), so there is no escalation, stability score, company baseline or cross filing pattern across runs.
• No website or domain lookup: the SEC's filer records carry no usable website (empty on 20 of 20 company rows of a run on 2026-11-01), so a domain such as apple.com cannot be matched to a company; use the ticker or the CIK.
• No hour windows: windows are whole days on the SEC's calendar (daysBack, dateFrom, dateTo), because search and window rows carry no acceptance time unless each filing's index page is read; cut by acceptedAt after the run.
• No contact e-mail address in its requests: the User-Agent names Automation Craft and its Store page.
• No free first rows: every filing row is charged from the first one.
• No own proxy URLs: proxyUrls in proxyConfiguration stops the run with a free row.
• No alerts or delivery connectors of its own (Slack, e-mail, a database): use Apify's schedules, integrations and webhooks on the dataset.

About the data

The data is the SEC's public filing data as filed, read from EDGAR (filer records, the full-text search, filing index pages, documents and XBRL facts) with a declared User-Agent at half the SEC's published request rate. The SEC's website policy says its information may be copied or further distributed by users of the website without the SEC's permission. Document texts are the document's own text: HTML is converted to plain text the way a browser shows it (runs of spaces collapsed, table cells separated by a tab), and a plain text document keeps its spacing and columns (only line ends are normalised and runs of blank lines shortened to one). This Actor is an independent tool. It is not affiliated with or approved by the U.S. Securities and Exchange Commission (SEC).

API examples

Every run is also an SEC filings API call over the Apify API: send the input, read the rows.

curl (synchronous run, returns the rows):

curl -X POST "https://api.apify.com/v2/acts/automation_craft~sec-filings-scraper/run-sync-get-dataset-items?token=YOUR_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"formTypes":["8-K"],"eventCategories":["earnings"],"daysBack":3,"maxItems":50}'

Node.js with apify-client:

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor('automation_craft/sec-filings-scraper').call({
    searchQueries: ['"material weakness"'],
    formTypes: ['10-K', '10-Q'],
    daysBack: 30,
    maxItems: 100,
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();
for (const row of items) if (row.type === 'filing') console.log(row.companyName, row.form, row.filedAt, row.filingIndexUrl);

Python with apify-client:

import os
from apify_client import ApifyClient

client = ApifyClient(os.environ["APIFY_TOKEN"])
run = client.actor("automation_craft/sec-filings-scraper").call(
    run_input={"tickers": ["AAPL", "MSFT"], "formTypes": ["10-K", "10-Q"], "maxFilingsPerCompany": 4, "includeFinancialData": True}
)
for row in client.dataset(run["defaultDatasetId"]).iterate_items():
    if row["type"] == "filing":
        print(row["companyName"], row["form"], row["reportDate"], len(row["financialData"] or []))

FAQ

How can I find a company's SEC filings?

Put its ticker into tickers (or its CIK into ciks, or its exact listed name into companyNames): the run reads the company's whole filer record, newest first, back to 1994 for the oldest filers. Narrow it with formTypes and a date window, and set maxFilingsPerCompany when you give a list of companies.

How far back does SEC EDGAR go?

Company filer records reach back to 1994 for the oldest filers, and this Actor reads them to their end, older pages included. The full-text search and the date window cover filings from 2001 on.

Can I search the full text of SEC filings past the 10,000 result limit?

Yes, in the default newest first order: a search with 10,000 or more results is split by date and, for one large day, by form, and every part is read (a single form on a single day with 10,000 or more matches is named in the run summary instead). Relevance order reads at most 10,000 results and says so in the run summary.

Are SEC filings public, and can I reuse the data?

Yes. The SEC's website policy says its information may be copied or further distributed without the SEC's permission, and every row links back to the filing on sec.gov. Some filers are people (insiders, reporting owners): their names, addresses, phone and EIN are delivered as the SEC publishes them. You are responsible for a lawful use of personal data (GDPR included).

Will the SEC block my runs?

The SEC refuses requests from undeclared tools with an HTTP 403 page. This Actor sends a declared User-Agent and asks at most 5 times per second, half the SEC's published limit of 10. If the SEC refuses the run's own address three times in a row, the run moves to one Apify datacenter proxy session; if that is refused too, it stops with stoppedBy: source_unavailable and nothing more is charged.

Why does this Actor run with limited permissions?

It runs with Apify's limited permissions, the least privilege level: it reads its input and writes only its own run's dataset and key-value store, and it reads its dataset back after a platform restart so no filing is charged twice. It opens no named store and touches nothing else in your account.

Changelog

See the Changelog tab of this Actor (CHANGELOG.md in the source).

More data tools by Automation Craft

• SEC Form D Scraper: Offerings & Funding Leads
• Yahoo Finance Scraper: Quotes, History, Financials
• US New Business Registrations Scraper - LLC Leads
• Google News Scraper: Search, Topics, Decoded URLs
• UK Contracts Finder Scraper: Tenders & Awards
• Grants.gov Scraper: Federal Grant Opportunities