A new problem-oriented cross-cultural method has been developed to identify the appropriate management of mental health problems in primary-care settings. It is designed to be used by multipurpose health workers in out-patient clinics in many developing countries. Presenting mental health problems in many cultural settings can be placed into one or more of eight behaviourly determined categories. After this initial categorization an algorithmic process is used to determine the sequence of observations which must be made in order to identify the appropriate management. The system allows for variations in available resources and local constraints. We outline the design principles of this algorithmic method and the rules for using it, and describe a training programme developed to teach people to adapt it and use it effectively. The evaluation studies that have been done in different developing countries confirm that the method is acceptable and efficient, and gives management decisions with a high degree of repeatability.