Building a currency converter with free APIs

1. Pick the right data source

For prototypes, Frankfurter is a practical starting point because it is simple and includes historical reference rates for many major pairs. For production, compare coverage, licensing, update frequency, uptime, historical depth, and whether the provider supports the currencies your users actually search for.

2. Normalize currency codes

Validate currency codes at the boundary. Do not let arbitrary strings move through your provider adapter.

const CURRENCIES = ["USD", "EUR", "GBP", "JPY", "CNY"] as const;
type Currency = typeof CURRENCIES[number];

function normalizeCurrency(value: string): Currency {
  const code = value.trim().toUpperCase();
  if (!CURRENCIES.includes(code as Currency)) {
    throw new Error(`Unsupported currency: ${value}`);
  }
  return code as Currency;
}

3. Separate the provider adapter

Your UI should not know whether a rate came from Frankfurter, a central bank file, a paid API, or your own database.

type FxRate = {
  base: Currency;
  quote: Currency;
  date: string;
  rate: number;
  source: string;
};

type RateProvider = {
  latest(base: Currency, quote: Currency): Promise<FxRate>;
  historical(base: Currency, quote: Currency, date: string): Promise<FxRate>;
};

4. Cache by pair and date

Historical daily rates should be cached by base, quote, and date. A rate for a past date rarely needs a fresh provider call.

async function getHistoricalRate(
  provider: RateProvider,
  base: Currency,
  quote: Currency,
  date: string
): Promise<FxRate> {
  const key = `fx:${base}:${quote}:${date}`;
  const cached = await cache.get<FxRate>(key);
  if (cached) return cached;

  const row = await provider.historical(base, quote, date);
  await cache.set(key, row, { ttlSeconds: 60 * 60 * 24 * 30 });
  return row;
}

5. Build pages users actually need

Strong exchange-rate pages should answer the next question, not just the first one. Include a current-rate hero, two-way converter, common amount rows, 1M/3M/1Y/5Y/MAX charts, 52-week high and low, related pairs, FAQ schema, and a clear reference-rate disclaimer.