A working-days API that gets the hard countries right

Most business-day calculators hardcode a Saturday–Sunday weekend. But Israel works on Sundays, the Gulf rests on Fridays, Indian banks close only on the 2nd and 4th Saturday of each month — and Nepal moved its weekend in April 2026. If your payroll run, settlement date or delivery SLA depends on a business-day count, a hardcoded weekend is a quiet source of wrong answers.

Same question, different answers

Every figure in the “our answer” column below is this API's live output, computed from statutory weekend rules and the country's real public-holiday calendar. The “naïve” column is what any Saturday–Sunday-hardcoded calculator produces for the identical question.

Question	Naïve Sat–Sun logic	Our answer	Why
Working days in Israel, 2–8 Aug 2026	5	6	Israel's weekly rest day is Saturday only — Sunday is a full business day. Source: Bank of Israel business-days publication.
Working days in India, August 2026 (whole month)	20	22	Banks close Sundays and only the 2nd & 4th Saturday (RBI, Gazette notification 20/08/2015 under s.25 NI Act). The 1st, 3rd and 5th Saturdays are full working days — and Independence Day (15 Aug) falls on one of them, which our holiday calendar catches too.
Next working day after Thursday 6 Aug 2026, Qatar	Friday 7 Aug	Sunday 9 Aug	Bank weekend is Friday+Saturday (QCB, Instructions to Banks, Part VII — banks' weekly holidays “are on Friday and Saturday”). A payment “due next business day” moves two days, not one.
Working days in Nepal, any week from April 2026	6 (stale data) if frozen on the old Saturday-only rule	5	Nepal Rastra Bank moved banks to a Saturday+Sunday weekend effective 6 April 2026. Our engine models the change as a dated rule period — historical queries still get the old rule, current ones get the new.

Check us, not our marketing. Every paid response carries the statutory instrument it rests on — source name, URL, effective date and when we last confirmed it. If a value can't be verified against a primary source, we serve null and say so. We'd rather refuse than guess: Iran's weekend is mid-transition and Somalia's is genuinely ambiguous, so for those we currently decline to answer.

Settlement dates: where a business-day count decides when money moves

The same arithmetic, applied to a trade date, answers “when does this settle?” — and this is where a wrong calendar costs real money. Every figure below is this API's live output.

Question	Naïve weekend-only count	Our answer	Why
US cash equity, T+1, traded Fri 9 Oct 2026	Mon 12 Oct	Tue 13 Oct	NYSE and Nasdaq trade on Columbus Day, but the Federal Reserve is closed and no money settlement occurs, so it is not a settlement day. Cycle cited to 17 CFR 240.15c6-1(a).
Japan, traded Thu 30 Apr 2026: equity vs government bond	one answer	Thu 7 May (equity) Fri 1 May (JGB)	Same country, same trade date, six calendar days apart: equities settle T+2, JGBs T+1, and Golden Week sits in between. An API that keys settlement off the country alone cannot express this.
UK, traded Thu 24 Dec 2026: gilt vs equity	one answer	Tue 29 Dec (gilt) Wed 30 Dec (equity)	Gilts settle T+1 by DMO market convention, not the T+2 that applies to equities under assimilated UK CSDR Art. 5(2).
Nigerian equity, T+1, traded Wed 30 Sep 2026	Thu 1 Oct	Fri 2 Oct	1 October is Independence Day. Nigeria also moved to T+1 on 1 June 2026 — we resolve the cycle by trade date, so trades either side of the switch get their own regime.

What we refuse to answer. US Treasuries settle T+1 by market convention — but no source we could obtain actually states that cycle, and the SEC rule that mandates T+1 expressly exempts government securities, so we refuse the question rather than serve a number we believe but cannot cite. Nigerian and Kenyan government bonds do not settle where their equities do (CBN/FMDQ and the CBK's DhowCSD, not CSCS or CDSC), and South African bonds have no single mandatory cycle — so we return a free refusal explaining that, rather than quietly reusing the equity cycle. We refuse when a calculation would run past the end of a published holiday calendar, because extrapolating next year's holidays is guessing. And we do not compute FX spot value dates at all: the market's USD-holiday test admits more than one documented reading, and the readings give different dates. Every cycle we do hold, with its citation, is published free at /settlement-conventions.

One honest boundary. These calculations run on national statutory holiday calendars and researched statutory weekend rules — not on the calendar of any CSD, clearing house or exchange. Those diverge in both directions, and every answer names the basis it used and the divergence we know about, so you can check us rather than trust us.

How this compares to the usual options

We reviewed the public documentation of popular commercial working-days APIs (checked July 2026). The pattern: no disclosed country coverage, no cited sources for weekend or holiday data, and an optional “custom weekend days” parameter — which quietly makes correct weekends your problem. We publish our rules, cite every value, and flag data past its verification window as stale. Run the four questions above against any calculator you're considering and compare.

Try it

# Free: what exists
curl https://afriref.dev/catalog

# Paid ($0.02, no signup): India, August 2026
curl -X POST https://afriref.dev/v1/answers/working-days \
  -H "Content-Type: application/json" \
  -d '{"country":"in","from":"2026-08-01","to":"2026-08-31"}'

# Paid ($0.05): when does a US equity trade struck on 9 Oct 2026 settle?
curl -X POST https://afriref.dev/v1/answers/settlement-date \
  -H "Content-Type: application/json" \
  -d '{"trade_date":"2026-10-09","calendars":["us"],"market":"us","instrument_class":"cash_equity"}'

# Free: every settlement cycle we cite, and what we refuse to cite
curl https://afriref.dev/settlement-conventions

Machine clients pay per call (HTTP 402 quotes the price and payment options). Humans and teams can buy prepaid credit from $10 — one-time purchase, no subscription, key delivered instantly.

Coverage

119 countries across seven regional services, with statutory weekend rules individually researched — including the Friday+Saturday Gulf and Levant, Saturday-only Israel, nth-Saturday India, and dated rule changes like Nepal 2026. Public-holiday calendars are served per country and every computed answer lists which holidays it applied. Sub-national exceptions we can't yet model (Malaysia's Fri+Sat states, Sharjah's three-day weekend) are stated as caveats on the answer rather than silently ignored.