Every claim traces back to a specific filing — ticker, form, date, section, and a link to the primary source.
FilingsIQ retrieves directly from 10-K, 10-Q, 8-K, DEF 14A, 20-F, and 6-K filings and cites the exact excerpt behind every claim. This is retrieval, not recall — if the filings don't say it, the answer says so instead of guessing.
Every claim is verifiable
Answers are built from the actual filing text, not model memory — each one maps back to a specific excerpt you can check yourself. (Hybrid vector + keyword search with cross-encoder reranking, for the technically curious.)
Auto-refreshing corpus
A recurring pipeline ingests new filings across NYSE and Nasdaq automatically. No one is pasting documents in by hand.
Honest about staleness
Ask something time-sensitive and the freshest filing on hand is old for that form type — the answer says so explicitly.
Handles compound questions
A question spanning two topics — say, revenue trends alongside a specific risk — gets broken into parts, researched separately, and brought back together in one answer.
What SEC filing types does it cover?
10-K and 10-Q (annual/quarterly reports), 8-K (material events, including earnings releases), DEF 14A (proxy statements — executive comp, governance), and 20-F/6-K (the equivalents for foreign private issuers like TSM or ASML).
Does it cover pre-IPO companies?
Selected Form S-1 registration statements are indexed — ask about a covered pre-IPO company's disclosed risks, use of proceeds, or business description. Answers drawing on an S-1 are flagged explicitly: a prospectus is the company's own sales document, not an audited annual report.
Does it include earnings call transcripts?
No — SEC filings never include the earnings call transcript itself, only the earnings release (8-K Item 2.02). That's a hard ceiling on what any SEC-filings-based tool can surface for "what did management say on the call."
How is this different from asking ChatGPT or Claude directly?
Every answer is grounded in retrieved filing excerpts with citations to ticker/form/section/source URL, instead of a model recalling training data. The corpus refreshes automatically and flags staleness explicitly.
Can I use this inside Claude Desktop or another AI assistant?
Yes — it's available as an MCP (Model Context Protocol) server exposing ask_sec_analyst, search_filings, and list_coverage tools. See the Developers page for setup.
How current is the data?
The corpus refreshes automatically on a recurring schedule, every answer cites the filing date of each excerpt it draws on, and if the freshest filing on hand is old for what you asked, the answer flags that explicitly. Developers can query exact per-company coverage via the API.
Is this free to use?
This is a demo/portfolio deployment, not a commercial product with an SLA — it may be rate-limited or occasionally unavailable.