Objective: To develop a semi-quantitative FFQ and to evaluate its validity and reproducibility for the assessment of total dietary intake of Kenyan urban adult population, given its non-existence in Kenya. Design: The current study adopted a cross-sectional design. A culture-sensitive semi-quantitative FFQ was developed and its validity was tested relative to three non-consecutive 24-h recalls (24hR). Reproducibility was tested by the test–retest method, with a 3-week interval. Spearman’s correlation coefficients and intra-class correlation coefficients were calculated for several macro- and micronutrients. Cross-classification into quartiles and Bland and Altman plots were analysed. Setting: Nairobi county (Dagoreti South and Starehe constituencies). Participants: A convenient sample was recruited in three different clusters in Nairobi. Results: A culture-sensitive 123-food-item semi-quantitative FFQ showed higher nutrient intakes compared with the 24hR (total energy median 12543·632 v. 8501·888 kJ, P < 0·001). Energy-adjusted and deattenuated Spearman’s correlations for macronutrients ranged between 0·21 (total fat) and 0·47 (protein). The agreement in the same quartile varied from 28 % (protein) to 41 % (carbohydrates). Including adjacent quartiles, the range increased: 76 % (protein and fat) to 81 % (carbohydrates). The extreme disagreement was low. The first FFQ application resulted in higher mean values for all nutrients compared with the second FFQ (total energy median 12459·952 v. 10485·104 kJ, P < 0·001). Energy-adjusted correlations for macronutrients ranged from 0·28 (carbohydrates) to 0·61 (protein). Intra-class correlation coefficients for macronutrients were moderate, between 0·6 and 0·7. Conclusions: The developed semi-quantitative FFQ was shown to be a valid and reproducible tool for ranking urban adult Kenyans according to their dietary intake.