Finance.Foundation
Data releases · current: seed-0.7 · marketcap-1 · 2026-08-30

Changelog

Every dataset release, dated and versioned. The API reports its version in every response as data_version.

Entries describe what shipped in that release and are not rewritten afterwards, so an older entry may state a cadence or a count that has since changed. The current state always wins: the live macro layer refreshes every three hours, delayed equity quotes rotate a batch every twenty minutes, World Bank indicators refresh daily, and the live counts are on /data.

2026-10-08

tokenization-1

  • Five depositories now carry the blockchain networks their own documents name for tokenised securities: KSD (Korea), DTC (United States), JSCC (Japan), and NSDL and CDSL (India). Each line has the document, its date, a status — depository guideline, pilot or planned — and the day it was checked against the document.
  • It is on each depository's page and in /api/v1/clearing as a tokenization field. No law in the sixteen jurisdictions read names a network, so the field records the guideline or release that does, never the statute.
  • Where the documents name no network the field says so rather than guessing: India's Demat 2.0 pilot runs on a ledger SEBI says the depositories own, without naming it.
2026-10-01

research-3

  • Research note #9, 'Balance Sheets Bigger Than Nations' — bank total assets summed by country against GDP, computed from the site's own banks and countries tables.
2026-09-02

quotes-off

  • The delayed equity quote layer is gone. It carried a closing price and daily change for 77 of 121 companies, shown on the company page and served at /api/v1/quotes, and it was read from Twelve Data's free tier.
  • The reason is Twelve Data's own agreement, which defines Redistribution as "any publication, distribution, or provision of Data to third parties" and permits it only under a paid add-on. A public page and a keyless public endpoint are both exactly that, whether or not anyone is charged for them. The free tier reported itself as plan_category "basic", 8 credits a minute and 800 a day, which is the $0 plan on their price list.
  • Nothing computed changed. Market capitalisation never passed through this layer: it is shares outstanding from the issuer's filing times the closing price on its home listing, and both inputs and their dates are still on every company page. What is lost is freshness on those 77 pages — the price shown is now the dated snapshot the capitalisation is built from, and nothing more recent.
  • The 20-minute cron that rotated the quotes is removed; the 3-hourly macro and markets cron is untouched. The 77 stored quotes and 255 rows of quote history were dropped from the database rather than left in place, and Twelve Data's entry left /sources.
  • No open replacement exists. Five free equity price sources were read against their own terms — Alpha Vantage, Financial Modeling Prep, Finnhub, Stooq and Twelve Data itself — and all five prohibit publishing their values to third parties; two are stricter than the one removed. The AWS Open Data registry holds one equity-market dataset out of a thousand, and it is deprecated. Exchange data is a licensed product, so this is a structural absence rather than a gap we failed to fill.
2026-09-01

research-2

  • Research note #8, 'The Shape of "Live" Data' — the latency and provenance structure across the 29 series behind /api/v1/live, computed from the feed itself.
2026-08-30

marketcap-1

  • Market capitalisation is now computed rather than stored. Every company carries a share count read from its own filing — SEC cover pages for 76 issuers, home-market disclosures (regulator share-capital declarations, exchange monthly returns, interim reports) for the rest — multiplied by the close on its primary listing, converted at the ECB reference rate. Both inputs, their dates and a link to the filing are published on each company page and in the API's new market_cap object.
  • The correction is large because the stored values had gone stale: 58 of 121 companies moved by more than 20%, 13 by more than half. The superseded hand-entered figure is kept as market_cap_seed_usd_bn so the two can be compared.
  • Nine companies are now priced per listed line rather than at a single blended price — A/H pairs (ICBC, Ping An, BYD, CATL, China Mobile), common/preferred pairs (Samsung Electronics, Petrobras), Atlas Copco's A and B lines and Rio Tinto's dual-listed plc and Limited legs. Applying one price to the whole share count was wrong in both directions.
  • Share counts that are not simply the cover-page number now say why: Visa is on an as-converted Class A basis (its B and C classes convert at 1.5x and 4.0x), Berkshire is expressed in Class B equivalents, and Toyota, Sony, TotalEnergies and BP are stated ex-treasury. Honeywell's one-for-two reverse split of 29 June 2026 is reflected.
  • New sources on /sources: SEC EDGAR (public domain) for share counts, and home-market closing prices for the venues our quote provider does not cover.
2026-08-29

seed-0.6 · hubs-1

  • Assets expanded 30 → 56: 8 more equity indices (Russell 2000, S&P/TSX, ASX 200, Ibovespa, SENSEX, NIFTY 50, KOSPI, SMI), 5 commodities (palladium, aluminium, nickel, soybeans, coffee), 5 FX pairs, 4 government bonds (gilt, BTP, OAT, India 10Y) and 4 crypto (XRP, BNB, ADA, DAI).
  • Live series 12 → 21: the four new crypto tokens from CoinGecko and five new FX crosses from the ECB daily reference rates, same three-hourly cron and the same provenance rules.
  • Reference page per identifier standard: /identifiers/isin, /lei, /figi, /mic, /cusip, /sedol — format anatomy broken down position by position, check-digit algorithm, a worked example from our own data, FAQPage structured data and a link into the validator.
  • Sector hubs at /sectors and /sectors/{sector}: every company in a sector with its LEI, combined capitalization, venue and jurisdiction spread, index exposure and the funds holding it.
2026-08-29

programmatic-1

  • Identifier reverse lookup: /isin/{isin}, /lei/{lei}, /figi/{figi}, /mic/{mic} resolve straight to the entity; unknown-but-valid codes get an honest 'not in our covered set' page pointing at GLEIF/OpenFIGI instead of a dead end. Segment MICs (e.g. /mic/XNGS) get their own page explaining the operating-vs-segment hierarchy.
  • Identifier crosswalk table on every company page: Ticker ↔ MIC ↔ ISIN ↔ LEI ↔ FIGI ↔ CUSIP, each linked to its lookup. CUSIP is derived from the US/CA ISIN and check-digit verified — validated, never redistributed as data.
  • Comparison pages: 431 pairs at /compare/{a}-vs-{b} for companies (market cap, venue+MIC, jurisdiction, identifiers, shared indices and fund holders) and venues (operator LEI, segment MICs, sessions, settlement, covered listings), plus a /compare index.
  • Embeddable cards: /embed/company/{id} and /embed/market/{id} — one iframe, dark-mode aware, noindex + canonical to the entity page, with CC BY 4.0 attribution and a link home. Docs at /embed.
  • Standards cheat sheets: copy-paste ISIN (mod-10), LEI (ISO 7064 mod-97) and MIC validation in Python and TypeScript.
  • /llms-full.txt — the complete entity inventory in one flat text file for LLM ingestion; new sitemap segments for compare and identifier pages.
2026-08-29

seed-0.5 · markets-2

  • Exchange pages deepened: 71 segment MICs across 19 venues (verified line-by-line against the official ISO 10383 registry, not from memory), trading sessions with lunch breaks, pre/post-market and settlement cycles, benchmark indices linked to asset pages, and operator LEIs resolved via GLEIF (18/20; Shenzhen and Nasdaq Stockholm left null rather than guessed).
  • Data correction: BME's operating MIC is BMEX — XMAD (Bolsa de Madrid) is one of its segments, not the operating code. Fixed.
  • Markets catalogue: filter chips by region (Americas/EMEA/APAC), instrument class (derived from each venue's own segment MICs, not guessed) and currency; ItemList JSON-LD with MIC and operator leiCode; cURL/Python/TypeScript snippet that mirrors the current query.
2026-08-29

llm-1

  • Markdown twins for every page: append .md (e.g. /markets/nyse.md) or send Accept: text/markdown — clean tables of identifiers, profile and graph edges for LLM ingestion. Listed in /llms.txt.
  • Identifier validator on /identifiers: checks real check digits in the browser for LEI (ISO 7064 mod-97), ISIN, CUSIP, SEDOL and FIGI, plus MIC and ISO 4217 formats. Licensed identifiers (CUSIP, SEDOL) are validated but never republished.
2026-08-28

ingest-1

  • Ingestion agent: POST /api/v1/ingest and MCP tool ingest_text turn unstructured financial text into FF-Schema entities and relationships (Workers AI extraction + strict validation: sector/asset taxonomies, relationship types, ISIN checksums, LEI/MIC formats; unresolved identifiers reported, never invented).
  • Read-only by design: extracted output is returned to the caller and is never written into the Finance Foundation dataset. Fair use: 20 requests/day per IP, 6000-char input limit.
2026-08-28

sponsor-1

  • Sponsor-a-dataset path on /data: support funds dataset depth and upkeep, credited with attribution — access is never gated. Free tier and the open MCP server stay full; paid tiers remain roadmap, to be built only on sustained API/MCP usage growth.
2026-08-28

research-1

  • Research note #7, 'The Concentration Graph' — the first of a monthly cadence, computed from the site's own dataset (country/company/fund-overlap concentration + the identifier-gap finding).
  • RSS feed at /rss.xml for research notes; <link rel=alternate> on every page.
2026-08-28

seed-0.4 · graph-2

  • New relationship types: member_of (index membership — covered subset per index: S&P 500, Nasdaq-100, DAX, EURO STOXX 50, FTSE 100, Nikkei 225, Hang Seng, CSI 300) and holds (top fund holdings among covered entities, approximate).
  • Interactive knowledge-graph map at /graph — every entity and edge, canvas-rendered, click-through to entity pages.
  • GET /api/v1/graph/full — the entire graph in one payload; neighborhood graphs and the MCP graph tool now include the new edges.
2026-08-28

mcp-1

  • Native MCP server at /mcp (Model Context Protocol, Streamable HTTP, stateless, no auth) — six tools: lookup, list, search, resolve_identifier, graph, live_data.
  • Connect from Claude Code: claude mcp add --transport http finance https://finance.foundation/mcp — or as a custom connector in Claude.ai/Desktop.
2026-08-28

live-1

  • First live data layer: 12 daily-updated series (BTC/ETH/SOL/USDT/USDC prices, EUR/USD·USD/JPY·GBP/USD·USD/CNY, ECB main refinancing rate, US 10Y Treasury yield, US CPI YoY) from keyless open sources (CoinGecko, ECB, US Treasury, OECD), stored in D1 with per-observation provenance.
  • New: GET /api/v1/live; live blocks on matching asset pages and /economy cards — labeled 'daily reference, not real-time'.
  • Euro-area HICP deliberately deferred: Eurostat/ECB discontinued the ICP dataset in Feb 2026; will attach the successor dataset when published.
2026-08-28

seed-0.3

  • Coverage expansion: 121 companies (was 34), 20 markets (+Shenzhen, ASX, B3, BME, Borsa Italiana, Nasdaq Stockholm, BSE), 21 countries, 30 assets, 25 funds, 20 banks.
  • Identifiers extended: LEI 119/121 (Chevron and Kweichow Moutai unverifiable — left null), ISIN 108/121 (check digit validated), FIGI 107/121 (name cross-checked).
  • Cross-site knowledge graph: bank entities now link their infrastructure profile on banking.foundation.
  • Datasets drafted by a 7-agent workflow and adversarially fact-checked before release (1 fix, 6 wording corrections applied).
2026-08-28

seed-0.2

  • External identifiers for all companies: LEI (GLEIF, hand-audited), ISIN (check digit validated, 33/34 — BlackRock ISIN deliberately null pending post-reorg verification), share-class FIGI (OpenFIGI, name cross-checked).
  • New: /identifiers resolver page and GET /api/v1/resolve?q= endpoint.
  • New: OpenAPI 3.1 specification at /api/v1/openapi.json.
  • This changelog.
2026-08-28

seed-0.1

  • Initial release: 34 companies, 13 markets, 15 assets, 8 funds, 10 banks, 16 countries, 8 indicators, 6 research briefs, FF Schema v0.1.
  • Open API v1 (collections, graph, search, schema, /all), llms.txt, sitemap.