OBJECTIVE: This study investigated associations between types and food sources of protein with overweight/obesity and underweight in Ethiopia. DESIGN: We conducted a cross-sectional dietary survey using a non-quantitative FFQ. Linear regression models were used to assess associations between percentage energy intake from total, animal and plant protein and BMI. Logistic regression models were used to examine the associations of percentage energy intake from total, animal and plant protein and specific protein food sources with underweight and overweight/obesity. SETTING: Addis Ababa, Ethiopia. PARTICIPANTS: 1624 Ethiopian adults (992 women and 632 men) aged 18-49 years in selected households sampled using multi-stage random sampling from five sub-cities of Addis Ababa. RESULTS: Of the surveyed adults, 31 % were overweight or obese. The majority of energy intake was from carbohydrate with only 3 % from animal protein. In multivariable-adjusted linear models, BMI was not associated with percentage energy from total, plant or animal protein. Total and animal protein intake were both associated with lower odds of overweight/obesity (OR per 1 % energy increment of total protein 0·92; 95 % CI: 0·86, 0·99; P = 0·02; OR per 1 % energy increment of animal protein 0·89; 95 % CI: 0·82, 0·96; P = 0·004) when substituted for carbohydrate and adjusted for socio-demographic covariates. CONCLUSION: Increasing proportion of energy intake from total protein or animal protein in place of carbohydrate could be a strategy to address overweight and obesity in Addis Ababa; longitudinal studies are needed to further examine this potential association.