Normalized SEC financial data,
priced per call, built for agents.

Why this API

Today's Buffett Quality leaders

Top 20 companies by Quality Score (0–100, blending return on invested capital, durability through the cycle, cash conversion, balance-sheet strength and capital discipline over a 10-year window; ties broken by revenue), recomputed every night from the latest SEC filings. Live from /spotlight — free.

A dash (–) means that component had no usable data for the company and was dropped, with the remaining weights renormalized — not a zero.

Endpoints

All endpoints are GET and return JSON. Companies are identified by SEC CIK number (free lookup: /company/lookup?q=AAPL): Apple=320193, Microsoft=789019, NVIDIA=1045810, Amazon=1018724, Alphabet=1652044, Tesla=1318605.

Available metrics for /aggregate and /screen: revenue, cost_of_revenue, gross_profit, operating_income, net_income, operating_cash_flow, investing_cash_flow, financing_cash_flow, fcf, capex, research_development, sga_expense, total_assets, total_liabilities, stockholders_equity, retained_earnings, long_term_debt, cash_and_equivalents, inventory, operating_lease_liability, gross_margin, operating_margin, net_margin, rd_intensity, current_ratio, debt_to_equity, roa, roe — and more (see /llms.txt).

Quickstart

1. See the price (any HTTP client)

Unpaid requests return HTTP 402 with x402 v2 payment requirements in the PAYMENT-REQUIRED header — including input/output schemas and worked examples for every route.

$ curl -i https://sec.multyplylabs.com/company/320193/financials

HTTP/2 402
payment-required: eyJ4NDAyVmVyc2lvbiI6Mi...   # base64: scheme, network, amount, payTo, schemas

2. Pay per call — Python

pip install "x402[evm]" eth-account

import asyncio
from eth_account import Account
from x402.client import x402Client
from x402.http.clients.httpx import x402HttpxClient
from x402.mechanisms.evm import EthAccountSigner
from x402.mechanisms.evm.exact import ExactEvmClientScheme

account = Account.from_key("0x...")          # wallet holding USDC on Base
client = x402Client()
client.register("eip155:8453", ExactEvmClientScheme(EthAccountSigner(account)))

async def main():
    async with x402HttpxClient(client) as http:
        r = await http.get(
            "https://sec.multyplylabs.com/company/320193/financials?fy_from=2023")
        print(r.status_code)                  # 200 — paid $0.02 automatically
        print((await r.aread()).decode()[:200])

asyncio.run(main())

2′. Pay per call — TypeScript

npm install x402-fetch viem

import { privateKeyToAccount } from "viem/accounts";
import { wrapFetchWithPayment } from "x402-fetch";

const account = privateKeyToAccount("0x...");     // USDC on Base
const fetchWithPay = wrapFetchWithPayment(fetch, account);

const res = await fetchWithPay(
  "https://sec.multyplylabs.com/aggregate/gross_margin?industry=73&size=mega&fy=2025");
console.log(await res.json());                     // paid $0.05 automatically

3. Or use an API key (prepaid / fiat customers)

$ curl -H "X-API-Key: YOUR_KEY" \
    "https://sec.multyplylabs.com/company/320193/financials?fy_from=2024"

Want an API key with volume pricing? Email help@multyplylabs.com.

Worked examples

Apple's financials (real response, abridged)

GET /company/320193/financials?fy_from=2025

{"cik": 320193, "years": [{
  "fiscal_year": 2025,
  "revenue": 416161000000.0,
  "gross_profit": 195201000000.0,
  "operating_income": 133050000000.0,
  "net_income": 112010000000.0,
  "eps_diluted": 7.46,
  "total_assets": 359241000000.0,
  "stockholders_equity": 73733000000.0,
  "operating_cash_flow": 111482000000.0,
  "capex": 12715000000.0,
  "fcf": 98767000000.0,
  "gross_margin": 0.4690,
  "operating_margin": 0.3197,
  "net_margin": 0.2691,
  "rd_intensity": 0.0830,
  "roe": 1.5191, ...}]}

Industry benchmarks: software gross margins

GET /aggregate/gross_margin?industry=73&size=mega&fy=2025

{"metric": "gross_margin", "cohorts": [{
  "industry_group": "73", "size_bucket": "mega", "fiscal_year": 2025,
  "p10": 0.41, "p25": 0.55, "median": 0.67, "p75": 0.78, "p90": 0.85,
  "mean": 0.66, "n": 42}]}

Screening: profitable large caps

GET /screen?fy=2025&filters=[["operating_margin",">",0.25],["revenue",">",10000000000]]

{"fiscal_year": 2025, "count": 87, "companies": [
  {"cik": 320193, "name": "Apple Inc.", "tickers": ["AAPL"],
   "revenue": 416161000000.0, "operating_margin": 0.3197, ...}, ...]}

Peer comparison

GET /compare/1065280?fy=2025          # Netflix vs its industry cohort

{"company": {"name": "NETFLIX INC", "gross_margin": 0.489, ...},
 "cohort_percentiles": [
   {"metric": "gross_margin", "median": 0.41, "p75": 0.58, ...}, ...]}

Fresh filings with document links

GET /company/320193/filings?form=10-K&limit=1

{"cik": 320193, "filings": [{
  "form": "10-K", "filed": "2025-11-01",
  "document_url": "https://www.sec.gov/Archives/edgar/data/320193/..."}]}

Curated insights: who improved margins the most?

GET /insight/movers?metric=gross_margin&fy_from=2023&fy_to=2025&min_revenue=1000000000

{"insight": "movers", "count": 20, "results": [
  {"name": "Example Corp", "tickers": ["EXMP"],
   "start_value": 0.31, "end_value": 0.47, "delta": 0.16}, ...]}

Curated insights: the R&D arms race

GET /insight/metric_trend?metric=research_development&agg=sum&fy_from=2015&fy_to=2025

{"insight": "metric_trend", "results": [
  {"fiscal_year": 2015, "value": 271.4e9, "companies_reporting": 3211},
  ...
  {"fiscal_year": 2025, "value": 702.9e9, "companies_reporting": 4187}]}

Quality Score: durable, capital-efficient businesses

GET /insight/buffett_scan?min_score=70&min_revenue=1000000000&limit=10

{"insight": "buffett_scan", "count": 10, "results": [{
  "name": "Example Corp", "tickers": ["EXMP"], "fiscal_year": 2025,
  "buffett_score": 87.9,
  "score_return_on_capital": 24.82,   // of 30: 10y median ROIC blended with industry percentile
  "score_durability": 20.0,           // of 20: worst-year ROIC, margin stability, revenue declines
  "score_cash_conversion": 20.0,      // of 20: FCF vs net income, FCF margin
  "score_balance_sheet": 9.86,        // of 15: net debt / EBIT, interest coverage
  "score_capital_discipline": null,   // of 15: no data here -- dropped and renormalized, not zeroed
  "score_basis": 85,                  // % of the 100-pt weight actually scored (coverage * 100)
  "roic_median": 0.1573, "roic_worst_year": 0.1471,
  "industry_percentile": 1.0, "peer_group_size": 12,
  "window_years": 10, "coverage": 0.85, "methodology_version": 3,
  "flags": [], "revenue": 56164630077.0}, ...]}

Raw normalized facts with provenance

GET /company/320193/facts?concept=revenue&fy_from=2023

{"cik": 320193, "facts": [{
  "concept": "revenue",
  "concept_raw": "RevenueFromContractWithCustomerExcludingAssessedTax",
  "value": 416161000000.0, "unit": "USD",
  "fiscal_year": 2025, "fiscal_period": "FY",
  "form": "10-K", "filed": "2025-11-01", "quality_flags": ""}]}

How payment works (x402)

x402 is the open protocol — stewarded by the Linux Foundation, backed by Coinbase, Cloudflare, Google, Visa and Mastercard — that turns HTTP 402 into a working payment flow for machines:

• 1. Request — your client calls any endpoint. The server replies 402 with payment requirements: $ amount, USDC contract, our address, network (Base, eip155:8453).
• 2. Sign — your client signs an EIP-3009 USDC transfer authorization for the exact amount. No gas needed on your side; payments settle in ~2 seconds.
• 3. Retry — the request is retried with the signed payment header. The Coinbase CDP facilitator verifies and settles it on-chain, and the data comes back with a settlement receipt in the X-PAYMENT-RESPONSE header.

Client libraries handle all of this automatically — x402 SDKs exist for Python, TypeScript, Go and Rust. All you need is a wallet holding USDC on Base. Payments are final (no chargebacks); repeated identical requests are your responsibility to de-duplicate.

MCP server — plug into Claude, Cursor, or any MCP client

Give your AI assistant direct, self-paying access to SEC data. Tools: lookup_company (free ticker or name to CIK), company_financials, company_facts, company_filings, industry_aggregate, screen_companies, compare_to_peers, market_insight, buffett_scan, buffett_value, standard_score, insider_activity, institutional_ownership, short_pressure, company_signal, top_signals, and list_insights (free catalog).

Option A — Hosted endpoint (no install)

Point any MCP client at https://sec.multyplylabs.com/mcp (Streamable HTTP) and pass your API key as an X-API-Key header. Free tools (lookup_company, list_insights) work without a key; paid tools return payment guidance.

// Claude Desktop / Cursor / Cline
{"mcpServers": {"sec-data": {
  "url": "https://sec.multyplylabs.com/mcp",
  "headers": {"X-API-Key": "YOUR_KEY"}
}}}

Option B — Self-paying stdio server (pays per call in USDC)

Runs locally and pays each tool call from your wallet via x402 — no API key needed.

1. Download the server (single file, no package needed)

mkdir -p ~/mcp && curl -o ~/mcp/sec_data_server.py https://sec.multyplylabs.com/mcp-server.py
pip install "mcp>=1.10,<2" "x402[evm]" eth-account httpx

2. Add it to your MCP client

// Claude Desktop: ~/Library/Application Support/Claude/claude_desktop_config.json
// (Cursor, Cline, etc. use the same shape)
{"mcpServers": {"sec-data": {
  "command": "python3",
  "args": ["/Users/YOU/mcp/sec_data_server.py"],
  "env": {
    "SECPLATFORM_URL": "https://sec.multyplylabs.com",
    "BUYER_PRIVATE_KEY": "0x..."        // wallet with USDC on Base — pays per tool call
    // or instead: "SEC_PLATFORM_API_KEY": "..." for prepaid access
  }}}}

3. Restart your client and ask a question

Then ask: "How does NVIDIA's R&D intensity compare to other mega-cap semiconductor companies?" — the agent calls compare_to_peers and industry_aggregate, pays a few cents in USDC, and answers with sourced numbers.

Data & methodology

Source and freshness

• Primary source: SEC EDGAR bulk XBRL data (companyfacts + submissions), re-ingested nightly after the SEC rebuilds its archives (~05:00 ET).
• New filings are detected intraday from EDGAR's real-time feed every 10 minutes, with an end-of-day completeness pass against the daily index.
• Coverage: every XBRL-filing US public company — 20,057 filers, 123M+ facts, fiscal years 1993–present (dense from 2009, when XBRL became mandatory).

Normalization (what you're paying for)

• Raw XBRL tags are mapped to 61 canonical concepts via a priority-ordered candidate system (e.g. revenue ← RevenueFromContractWithCustomerExcludingAssessedTax, Revenues, SalesRevenueNet, ...), continuously expanded by breadth-of-use analysis.
• Comparative-period traps are resolved: a 10-K reports prior years under the same fiscal-year tag; we select the correct current-period fact by period-end, form priority and filing recency.
• Quality flags: implausible magnitudes (the $555M-tagged-as-$555B class of filing errors), unmapped custom extensions, and abnormal durations are flagged in /facts and excluded from aggregates.
• Derived metrics are computed consistently: margins, FCF, ROE/ROA, leverage and liquidity ratios from canonical inputs.
• Aggregates are cohort percentiles by 2-digit SIC × revenue size bucket × fiscal year, with a minimum cohort size of 3.

Quality Score (methodology v3)

• 0–100 over a 10-year window (5–9 years for companies with a shorter filing history) from five continuous components: return on invested capital (30 pts — median ROIC, blended 60/40 with the company's 2-digit-SIC industry percentile), durability through the cycle (20 — worst-year ROIC, operating-margin stability, revenue declines), cash conversion (20 — FCF vs net income, FCF margin), balance-sheet strength (15 — net debt / EBIT, interest coverage), and capital discipline (15 — diluted share-count change, net buybacks net of stock compensation).
• A component with no usable data for a company is dropped, not zeroed, and the remaining weights are renormalized; score_basis reports what fraction of the 100-pt weight was actually scored. A 60% floor exists as a safety net, but it currently excludes nobody — the lowest score_basis in the published set is 70.
• Not every company is scored. The exclusions that actually occur, with today's counts: stopped filing (4,313); fewer than five fiscal years on file (1,702); five or more years on file but too few of them yield a usable return on capital — missing EBIT tags, or an unclassified balance sheet, common among homebuilders and industrials with captive finance arms (1,612); and balance-sheet financials (banks, insurers, REITs, investment funds; 1,539) — capital is their inventory, not something returns are earned on. That leaves 2,885 of 12,051 companies — roughly a quarter of all filers, ~37% of active ones — scored today.
• Replaced an earlier equity-based version (owner earnings, ROE consistency, liabilities/equity, FCF quality) that scored high-ROIC, serial-buyback companies near zero once buybacks drove their book equity negative.

Valuation (Buffett Value Scan)

• Market cap = previous close x shares outstanding, refreshed nightly from EOD prices; every valuation row carries its close_date.
• Intrinsic value = 10-year DCF on normalized owner earnings (net income + D&A − capex, taken as the lower of the latest year and the 5-year median, so a cyclical peak isn't extrapolated for a decade), growth = 4-year OE CAGR capped at 10%, 12% discount rate, 2.5% terminal growth. The model pays ~9x owner earnings at zero growth and ~18x at the cap — deliberately conservative, so a large margin of safety is rare and means something.
• Value Score (0-100) = FCF yield (40) + margin of safety (40) + P/E (20). Owner-earnings yield is intentionally excluded: it is the same signal as the other two, and counting cheapness three times made every cheap stock score 100.
• Share counts come from the filing cover page and are split-adjusted: a split after the fiscal year end would otherwise pair post-split prices with pre-split shares and understate market cap. Only the common-stock ticker is priced — preferred series (e.g. XYZ-PD) are excluded.
• Sanity guards: rows implying a P/E under 2, a cash yield over 50%, or a sub-$10M market cap on a real operating business are dropped, not published — those are data errors, not bargains. Price coverage is checked each night and logged.
• Foreign private issuers (20-F filers — Chinese ADRs and similar) are excluded by default: they screen cheap for structural reasons fundamentals can't measure (VIE structures, delisting and governance risk). Detection uses the annual filing form, not state of incorporation — Arch Capital is Bermuda-incorporated but files a 10-K and belongs in results. Pass exclude_foreign=false to include them.
• Known limits: maintenance capex is approximated by full capex (understates asset-light compounders); balance-sheet financials (banks, insurers, REITs and investment funds) are excluded from the Quality Score entirely rather than scored, since capital is their inventory and return on invested capital isn't a meaningful measure for them; debt is sometimes an upper-bound proxy rather than a tagged figure, flagged as balance_sheet_estimated on the row.

Data is derived from public-domain SEC filings. Scores are heuristics inspired by Buffett's published writing — they are not Buffett's method, not a recommendation, and not investment advice. Prices are sourced from public EOD feeds and may contain errors.

FAQ

How do I find a company's CIK?

Call the free /company/lookup?q=AAPL with a ticker or company name (MCP: lookup_company), query /screen by criteria and read CIKs from results, or use the SEC's company search. Common ones: Apple 320193 · Microsoft 789019 · Amazon 1018724 · Alphabet 1652044 · Meta 1326801 · Tesla 1318605 · NVIDIA 1045810 · Netflix 1065280 · JPMorgan 19617 · Berkshire 1067983.

What do I need to pay with x402?

A wallet private key holding USDC on Base (the buyer needs no ETH — transfers are authorized off-chain and the facilitator pays gas). $1 covers 10–200 calls depending on endpoint.

Can I get quarterly data?

/facts includes quarterly (10-Q) facts today — filter by fiscal_period in the response. Clean quarterly statements and point-in-time snapshots are on the roadmap.

Do you have an SLA / higher rate limits / bulk pricing?

Email help@multyplylabs.com for API keys, volume pricing, or custom datasets.

Where else can my agent discover this service?

We're indexed in the Agentic Market / x402 Bazaar (via the Coinbase CDP facilitator) with machine-readable schemas in every 402 response, plus /llms.txt and /.well-known/x402.json.