Datasets
Central bank data API — statements, minutes, speeches
Central bank communication is the highest-value text in finance and the worst-organized. Statements live on one page, minutes appear weeks later on another, speeches are scattered across regional sites and PDFs, and projection rounds are published as spreadsheets whose format changes every few years. Anyone who has tried to diff two policy statements programmatically knows the whitespace alone will ruin an afternoon.
Hindcast is building a normalized corpus across the Fed, BCB, ECB, BoE, BoJ and the major LatAm banks: every document with its release timestamp, speaker, venue and document type, chunked for retrieval and diffable against the previous release of the same series. Projection rounds keep every vintage, so the question is not only what the committee projects but what it projected before.
For agents this turns hawk-or-dove into a query rather than a research project — with the underlying paragraph and its URL returned alongside any conclusion, so a claim can be checked instead of trusted.
hindcast — session
previewhindcast> diff the last two FOMC statements — what language changed?
→ tool: cb_statements { bank:"FED", series:"fomc_statement", diff:"previous" }
✓ paragraph-level diff with added, removed and reordered clauses
source: Federal Reserve press releases, both release timestamps retained
What's covered
- FOMC statements, minutes and press conference transcripts by release date
- BCB Copom statements and minutes, Portuguese source plus English
- ECB, BoE and BoJ policy statements and accounts of meetings
- Speeches with speaker, role, venue and delivery timestamp
- Projection rounds kept as vintages rather than a single current table
- Paragraph-level anchors so any quoted claim links back to the source
Join the waitlist — this page's dataset is already selected.
Questions
Are minutes aligned to the meeting or to their publication?
Both, and the distinction is load-bearing. Minutes describe a meeting that happened weeks before they were published; a point-in-time query returns them from their publication date, because that is when the market could read them.
Which LatAm banks are in scope?
Brazil is the deepest by design, with the other major regional banks planned alongside the majors. The catalog is built in demand order, so the vote on this page is the fastest way to move a specific bank up the list.
Can an agent quote from these documents safely?
That is the intent: every extract returns with its document, section and release timestamp, so an agent can cite the source rather than paraphrase from memory.
Related
Join the waitlist. Vote the roadmap. First in, first served.