SEC EDGAR Scraper

A company files an 8-K because the law gives it no choice, and what lands in EDGAR is the filed document itself rather than somebody's write-up of it. Two ways to get filings out: run a full-text search across everything the SEC has indexed, or pull one company's filing history from a ticker or CIK. Either way you get one row per filing with a direct link to the document.

No login, no API key. EDGAR is a public government system and this actor talks to it directly.

What you get

{
  "form": "20-F",
  "companyName": "Xiao-I Corp",
  "cik": "0001935172",
  "ticker": "AIXI",
  "filingDate": "2023-04-28",
  "reportDate": "2022-12-31",
  "accessionNumber": "0001213900-23-033683",
  "primaryDocument": "f20f2022ex4-19_xiaoicorp.htm",
  "description": "ENGLISH TRANSLATION OF AI CLOUD PLATFORM SERVICE CONTRACT BETWEEN SHANGHAI XIAO-I ROBOT TECHNOLOGY CO., LTD. AND CUSTOMER F DATED JUNE 27, 2022",
  "filingUrl": "https://www.sec.gov/Archives/edgar/data/1935172/000121390023033683/f20f2022ex4-19_xiaoicorp.htm"
}

filingUrl points at the actual document, not a landing page, so you can hand it straight to a parser or a summarizer.

Search mode returns thinner rows than company mode

Company mode gives you complete rows. Search mode doesn't.

The SEC's full-text index is sparse. It knows which document matched your phrase, but it often doesn't carry the ticker, the report date or the description alongside it. So search rows frequently have ticker, reportDate and description as null, while company rows are fully populated.

That's the SEC's index, not a gap in the scraper. If you need complete records, resolve your companies first and use company mode. If you're doing discovery — "who mentioned this phrase" — search mode is the right tool and the nulls don't matter.

Full-text search also won't page past 10,000 results for any one query. That's the SEC's own ceiling. If your phrase is common, narrow it with forms or a date range rather than expecting to page through everything.

Input

Find every filing mentioning a phrase:

{
  "mode": "search",
  "query": "artificial intelligence",
  "forms": "10-K,8-K",
  "startDate": "2024-01-01",
  "maxItems": 500
}

Pull filing histories by ticker:

{
  "mode": "company",
  "tickers": ["AAPL", "TSLA", "0000320193"],
  "forms": "10-K",
  "maxItems": 200
}

• query — matched as a phrase across filing documents. Required in search mode, ignored in company mode.
• tickers — tickers or numeric CIKs, mixed freely. Company mode only.
• forms — comma-separated form types, e.g. "10-K,8-K". Empty means all forms. Works in both modes.
• startDate / endDate — YYYY-MM-DD, both optional.
• maxItems — a total budget for the whole run, shared across every company and query. Not per company. Four tickers at maxItems: 100 does not give you 100 each; the first ticker can eat the budget before the fourth is reached. Default 100, ceiling 10,000.

Optionally mirror each filing into Notion with notionConnector.

Limits

• Full-text search: 10,000 results max per query, sparse metadata as described above.
• maxItems is a run total. Order your tickers accordingly.
• This returns filing metadata and a link. It does not download or parse the document text. Feed filingUrl to whatever does that.
• Company mode returns the recent filing history the SEC exposes per company, not an unbounded lifetime archive.

Proxy

Off by default and that's correct. SEC's public APIs have no anti-bot protection, so a proxy buys you nothing here and this actor does no IP rotation. Turn it on only if you want a specific geolocation or your own network is being rate-limited. The SEC does ask for a descriptive user agent and rate-limits politely; the actor handles that.

Billing

$2.00 per 1,000 filings, and there is no run-start fee.

You pay per filing row delivered. A search that matches nothing costs nothing.