We present the design and implementation of a cost-effective e-Health system for automatic depression detection. The system is based on a client-server architecture, where clients are popular mobile devices. For practical deployment, various factors that affect the accuracy and speed of depression detection are discussed and evaluated with extensive experiments.