Where do the exchange rates come from?

The free rail uses the European Central Bank’s daily reference rates via the Frankfurter API. These are official once-a-day reference values — published around 16:00 CET on ECB working days — not live dealing quotes.

The Pro rail adds keyed coverage from ExchangeRate-API: 160+ currencies and every pair. Whichever rail is active, the screen always tells you which provider the number came from and how fresh it is.

How does the converter calculate an amount?

Conversions are computed from the active rail’s rates for your exact pair. What you see is the mid-market reference value — the midpoint between buy and sell — which is the honest benchmark to compare any offer against.

Your bank or transfer service will apply its own margin and fees on top of the mid-market rate, so the amount you actually receive will differ. Use the reference value to judge how big that margin is.

Where does the bond-yield data come from?

US Treasury yields come from the FRED® API of the Federal Reserve Bank of St. Louis (daily) and US Treasury FiscalData. Other countries use monthly OECD long-term series. This product uses the FRED® API but is not endorsed or certified by the Federal Reserve Bank of St. Louis.

Yields matter to currencies because rate differentials drive capital flows: when one country’s 10-year yield rises against another’s, holding that currency pays more, which tends to support it. The yields dashboard shows those spreads directly.

How are the AI trend outlooks computed?

Each outlook starts from a deterministic signal — no AI involved: recent momentum (the pair’s ~30-day move in its daily ECB rate history) plus carry (the 10-year government-yield spread between the two currencies), combined into a composite that sets the direction and a confidence between 50% and 90%. Same inputs, same signal — every time.

Outlooks are precomputed nightly for every pair among the eight currencies with a yield rail behind them (USD, EUR, JPY, GBP, CHF, AUD, CAD, KRW) — 56 pairs in total. Other pairs show clearly-labeled sample content instead.

What does the AI actually write?

The short outlook text is written by an AI model (Anthropic Claude) from the computed facts only — the pair, its rate history, the yield spread and the signal. The model explains those facts; it is not asked to predict, and no personal data is involved.

The direction, strength and confidence you see on screen always come from the computed signal itself, never from the text. The prose is commentary; the numbers are the source of truth.

What do the data badges mean?

Every screen tells you what you are looking at: how fresh the number is and which provider produced it. If you are offline, the app shows the last rates it stored on your device — labeled as cached, never passed off as live.

Sample data — bundled demonstration content, like the outlook teaser for pairs without a yield rail — is always badged as sample. You are never shown sample data unlabeled.