Foreign exchange (FX) dealers are exposed to currency risk through both market and counterparty activities. Research in FX risk management has mainly focused on long-term risks, yet trading costs associated with long-term strategies make them undesirable for short-term risk hedging. In this paper, a short-term risk management system for FX dealers is described, in which the optimal risk-cost profiles are obtained through dynamic control of the dealer's positions on the spot market. This approach is formulated as a stochastic receding horizon control (SRHC) problem, incorporating elements that model client flow, transaction cost, market impact, exchange rate volatility and fluctuations caused by macroeconomic announcements. The proposed technique is backtested using both synthetic and historical client trade data. The results obtained outperform three benchmark hedging strategies on a risk-cost Pareto frontier, achieving up to a 47.6% cost improvement over benchmark strategies. A flexible scenario generation oracle is also introduced and used to quantify the effects of predictive model quality on risk management.