SEC EDGAR API · filing search · MCP + REST

SEC filings you can search and pull in one call — cited passages, not PDFs

An SEC EDGAR API should not hand you a 200-page HTML file. Ask a question — “what did Amazon say about commitments not yet recognized?” — and get the exact passage from the 10-K, 10-Q or 8-K, with the accession number and the sec.gov link, in one call. Connect the same search to Claude, Cursor or any MCP client and the model reads the filings for you, and cites them.

Get a free API key Search filings from your LLM

Free calls to start on your first key, no card. Then credits — no subscription.

One call. The passage, the form, the accession number, the link.

A real query and the real answer, trimmed to the fields that matter.

REST · search a filing

curl -s https://mcp.evidinvest.com/v1/tools/search_sec_filings \
  -H "Authorization: Bearer $EVIDINVEST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"arguments":{"query":"commitments not yet recognized","ticker":"AMZN"}}'

Response · Amazon Q2 2026 10-Q · trimmed

{
  "results": [
    {
      "text": "...those commitments not yet recognized were approximately $496 billion as of June 30, 2026. The weighted-average remaining life of our long-term contracts is 6.4 years...",
      "form_type": "10-Q",
      "period_of_report": "2026-06-30",
      "accession_number": "0001018724-26-000026",
      "citation": "AMZN 10-Q 2026-07-31 (Item 1)",
      "source_url": "https://www.sec.gov/Archives/edgar/data/1018724/000101872426000026/amzn-20260630.htm",
      "confidence": 0.91
    }
  ]
}

Open the URL and read the paragraph yourself. That is the whole contract: nothing comes back without a document behind it.

text: The exact section text — the passage, not a summary of it.
form_type · period_of_report: Which filing (10-K, 10-Q, 8-K, press exhibit) and the period it covers.
accession_number: The EDGAR accession number, so the passage can be audited.
citation: A one-line human citation: ticker, form, filing date, section.
source_url: The sec.gov document the text was taken from.
confidence: How well the passage matches the question; low scores are returned, not hidden.

What can be searched

Three corpora, all built from EDGAR documents. Each result names the tool and the filing it came from.

Filing sections
10-K, 10-Q and 8-K text by section — Item 1A risk factors, MD&A, notes, M&A and contract terms — plus the earnings press exhibits attached to 8-Ks.
search_sec_filings
Earnings calls and press exhibits
Verbatim quotes by speaker turn: prepared remarks, Q&A, and the same-day press-release exhibit, with speaker, role and date.
search_earnings_transcripts
Supply chain and risk, cited
Structured profiles built from filings: products, suppliers, customers, strategic priorities; supplier/customer edges with the verbatim filing quote; Item 1A concentration risks; chokepoint companies cited by the most other filers.
sec_get_profile · sec_get_supply_chain · sec_get_risks · sec_find_chokepoints · sec_compare

REST · search an earnings call

curl -s https://mcp.evidinvest.com/v1/tools/search_earnings_transcripts \
  -H "Authorization: Bearer $EVIDINVEST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"arguments":{"query":"capital expenditures outlook","ticker":"META"}}'

Looking for the numbers instead — statements, growth rates, prices, valuation ranges? Those are on the financial data page; same key, same credits.

Search the filings quickly with an LLM connected

Add the MCP server once. From then on your model can search filings and calls itself, and every answer it gives you carries the citation it found.

Claude Code · one command

claude mcp add --transport http evidinvest https://mcp.evidinvest.com/mcp \
  --header "Authorization: Bearer YOUR_API_KEY"

claude_desktop_config.json · .cursor/mcp.json

{
  "mcpServers": {
    "evidinvest": {
      "type": "streamable-http",
      "url": "https://mcp.evidinvest.com/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}

Streamable HTTP, bearer key. The same 48 tools are exposed to the model as the REST API — filing search, transcripts, statements, valuation, supply chain.

What a cited answer looks like

You ask

“What did Meta say about full-year capex in its latest filing, and where?”

The model answers, with the passage it found

Meta expects full-year 2026 capital expenditures, including principal payments on finance leases, to be in the range of $130–145 billion, narrowed from a prior outlook of $125–145 billion. Source: META Q2 2026 press exhibit, 8-K, accession 0001628280-26-050596, filed 2026-07-29 — sec.gov/Archives/edgar/data/1326801/000162828026050596/meta-06302026xexhibit991.htm.

The wording is the model’s; the figures, form, accession number and link are copied from the search result. You can check every one against sec.gov.

01
Create a key
Sign in, open API keys, create one, copy it once. Free calls to start, no card.
02
Connect your client
Claude Code, Claude Desktop, Cursor or any MCP client — one config block. Or call the REST endpoints directly.
03
Ask, then verify
Every passage the model quotes carries its accession number and sec.gov link. Open the link when it matters.

You get more when you connect your own model. A search from your LLM only spends credits on the filing data, not on our model writing an analysis — so the same credit pack goes much further than it does on the website.

Why not EDGAR full-text search or a filing feed?

Both are free and both are useful. Neither returns a passage you can hand to a model or a colleague as evidence.

Point	EDGAR full-text search	Filing feed / alert service	EvidInvest filing search
What you get	A list of documents that contain the words.	A notification that a document exists.	The passage that answers the question, with its citation.
Meaning, not just keywords	Exact words only.	None.	Semantic search over sections and speaker turns — “commitments not yet recognized” finds the note even when the wording differs.
Earnings calls	Not covered.	Rarely.	Prepared remarks, Q&A and press exhibits, quoted verbatim by speaker.
From an AI agent	Only if you write the scraper.	No.	Native MCP server; the model calls the search itself.
Cost	Free, plus your time.	Subscription.	Free calls to start, then credits that never expire.

Questions developers ask

Which forms are indexed?: 10-K, 10-Q and 8-K sections, including the earnings press exhibits filed with 8-Ks, plus earnings-call transcripts by speaker turn. Structured supply-chain and risk profiles are built from the same filings.
How fresh is it?: Earnings exhibits and new filings are indexed as they are ingested from EDGAR — same-day for earnings exhibits in the cases we check most. If a document is not yet indexed, the search returns nothing rather than a guess.
Can I get the whole filing?: Every result carries the sec.gov URL of the source document, so the whole filing is one click away. The API itself returns sections and passages, which is what a model or a footnote needs.
What does it cost?: Free calls to start on your first key, then usage is metered against a credit balance — packs from $10, shared with the website and MCP, no subscription. 60 requests per minute per key.

Get a free API key MCP setup guide

EvidInvest never says buy or sell. It shows what the filings say and how each number was computed. Research, not investment advice.