SEC EDGAR filing research

​

Filing research shape

• Browser-agent research against a public, source-of-record site.
• Domain policy for a single trusted origin.
• Structured fields that an application can validate before storage.
• Source URLs as evidence, not final authority.

​

Setup

​

Run it from the CLI

web agent \
  --url https://www.sec.gov/edgar/search/ \
  --allow-domain sec.gov \
  "Find Apple's latest 10-Q filings. Return the first 5 filings with form type, filing date, accession number, filing detail URL, and a one-sentence summary. Do not download bulk data or submit forms beyond public search."

​

Expected output

Apple Inc. 10-Q filings

1. Form: 10-Q
   Filing date: ...
   Accession number: ...
   URL: https://www.sec.gov/Archives/...
   Summary: ...

​

Build it into an app

const result = await agent.run({
  startUrl: "https://www.sec.gov/edgar/search/",
  goal:
    "Find Apple's latest 10-Q filings. Return the first 5 filings with form type, filing date, accession number, filing detail URL, and a one-sentence summary.",
});

if (result.status !== "completed") {
  await queueReview(result);
}

​

Production notes

• Domain policy: sec.gov.
• Output fields: form type, filing date, accession number, detail URL, summary.
• Use structured output before writing to a database.
• Persist source URLs with the result.
• Treat SEC pages as public evidence that can change over time.

​

Inspect

• Debug UI when a search result looks wrong.
• Recording for replayable evidence.
• Steps for page titles and URLs.
• Status for blocker or timeout state.

​

Cleanup

​

Common failures

• A rate limit, bot challenge, or temporary SEC page error should route the run to review instead of inventing filing data.
• If the site layout changes, inspect the recording and tighten the prompt around visible labels or detail-page URLs.