SweetValueLab
Write for us

Open datasets

The tables behind the boards, the screen and the gauge, as CSV and JSON. Rebuilt every night with the site from the same data the pages show; free to reuse with attribution.

US stocks: price and model value

nightly

One row per covered stock: last close, the median model value, the gap between them, the snapshot's quality flag, and the reverse-DCF inputs (market value, net debt, free cash flow, P/E, five-year FCF growth). The screen at /stocks, as a table.

Reverse DCF inputs

nightly

One row per covered stock: the base free cash flow, enterprise value and history the reverse DCF starts from, with EV/FCF. The implied growth depends on the reader's required return and is solved on each stock's page.

Bank Health Board

nightly

One row per bank on the board: P/TBV and the return-driven theory line, ROTCE, efficiency, CET1, credit cost, AOCI against tangible equity, the three lights and every grade. Trailing twelve months from filed quarters.

REIT Board

nightly

One row per REIT on the board: Cash FFO (operating cash flow less working capital, standing in for funds from operations) and the multiple on it, dividend yield and Cash FFO payout, Cash FFO growth, net debt to EBITDA, interest cover, the lights and the grades.

Software Board

nightly

One row per software company on the board: revenue growth and its prior year, gross and free-cash-flow margins with and without stock-based compensation, SBC to revenue, share count change, EV multiples, the lights and the grades.

Market sentiment gauge history

nightly

One row per trading day: the composite score, its band, each factor's percentile, and the S&P 500 close. The gauge at /tools/market-sentiment over its trailing window.

Tracked 13F filers

weekly

One row per tracked manager: the quarter of the latest 13F on file, its filing date, the number of positions and the reported portfolio value.

Earnings calendar

nightly

One row per scheduled report in the calendar's window: date, company, sector, market value, consensus EPS and revenue, and the actuals once reported.

Quarterly releases

Archived with a DOI on Zenodo, mirrored on Hugging Face and Kaggle, and cited by version.

Superinvestor 13F holdings history

quarterly · version 2026-Q2

Every quarterly 13F book of the 82 tracked managers since Q1 2013: 227,970 positions with CUSIP, ticker, shares, value and the EDGAR filing each row comes from, amendments applied and values in US dollars. Taken from SEC EDGAR; the same books the investor pages show.

Zenodo ↗Hugging Face ↗Kaggle ↗DOI 10.5281/zenodo.23113678· from/investors

Use it from your AI assistant

The same numbers are a read-only MCP server, so an assistant such as Claude or Cursor can look them up while it answers and link the page it read them from. No key and no account.

https://api.sweetvaluelab.com/mcp
  • search_stocks — finds a covered stock by ticker or company name
  • get_stock_valuation — the price, the headline model value and the gap to it, every other model with its range, and the discount rate
  • get_market_context — the Treasury yield curve, the rates the models discount with, and the sentiment gauge

Claude (claude.ai or the desktop app): Settings → Connectors → Add custom connector, then paste the address.

Claude Code:

claude mcp add --transport http sweetvaluelab https://api.sweetvaluelab.com/mcp

Cursor and other clients, in the MCP config file:

{
  "mcpServers": {
    "sweetvaluelab": {
      "url": "https://api.sweetvaluelab.com/mcp"
    }
  }
}

It answers from the same nightly data as the pages: model output, not live quotes, and not investment advice. Listed in the official MCP Registry as com.sweetvaluelab/valuation.

How to use them

Every file is written by the nightly build from the same page data the site renders, so a row here is the row on the page. The JSON carries the provenance around the rows (asOf, the source page, the licence, the column list); the CSV is the rows alone, one header line, RFC 4180 quoting. Nested figures are flattened with underscores (grades_rotce, lights_rate_on); series and lists are left to the pages.

Licence: CC BY 4.0. Reuse, redistribute and build on them; credit SweetValueLab with a link. The figures are computed from public filings and market data and may be incomplete or wrong; the methods are described on each source page. Research and educational data, not investment advice.

Files are served with open CORS, so a script or a notebook can fetch them directly; the URLs are stable and the content changes once a night.