In this paper, we present a Markov-Chain Monte Carlo (MCMC) simulation method for valuing a European-style currency option in a regime-switching framework of exchange rate. Since the Bayesian inference via Gibbs sampling incorporates uncertainty associated with the underlying parameters of the regime-switching model, Our model generalizes the lattice-based regime-switching model by Bollen (1998). Using the sterling (to US dollar) Swiss franc, Australian dollar and Japanese yen, the Black-Scholes model is shown to generate significant pricing errors when a regime-switching process governs underlying asset returns. The computed predictive option pricing are shown to generate the implied volatility smiles commonly found in earlier empirical studies.