We evaluated the International Classification of Primary Care interobserver reliability for coding the complaints reported at a primary public health unit. The study sample consisted of 300 appointments. Reasons for appointment were registered by physicians on the medical record and coding was performed by two investigators. Reliability among the classifiers as to the number of coded reasons was calculated by weighted kappa. Crude kappa coefficients were used for the remaining analyses (chapter codes and full codes). A total of 634 reasons were coded by the first classifier and 699 by the second one. Crude reliability of numbers of reasons was moderate, with a 0.71 weighted kappa (95% Confidence Interval [95%CI]: 0.68-0.73). Regarding chapter codes and full codes, reliability was substantial (kappa = 0.89; 95%CI: 0.87-0.91) and moderate (kappa = 0.72; 95% CI: 0.68-0.76), respectively. Our results suggest a good performance of the Classification for coding reasons for visit in primary care, even when data are obtained from physician consultation notes.