query_company_filings

query_company_filings fetches a public US company’s most recent 10-K and returns extracted sections rather than the raw 500–900 KB markdown, so the response stays well under any client context cap. The full markdown is still materialised as a dataset for follow-up slicing.

When to use it

• The user asks about a public company’s risk factors, AI or cloud strategy, cybersecurity governance, or recent M&A.
• You ran account_research and want a section it didn’t surface by default.

Inputs

Parameter	Type	Required	Description
website	string	Yes	Bare company domain, e.g. "capitalone.com".
sections	string[]	No	Section ids to extract. Omit to extract all.
keywords	string[]	No	Ad-hoc keywords for extra ~500-character extractions.
max_chars	integer	No	Per-section cap. Default 4000, max 8000.
limit	integer	No	Filings to consider. Default 1 (most recent).

Section ids: cybersecurity, ai, cloud, ma, tech, financial, risk_factors.

Returns

JSON with the extracted sections, any keyword extractions, and a dataset handle over the full filing markdown.

found: false is returned when the company is not an SEC filer (common for non-US companies) — that’s not an error; the caller simply continues without 10-K data.

Example agent prompt

“What does Capital One’s latest 10-K say about cybersecurity governance and AI?”