OBJECTIVE: To assess the inter-observer variability and accuracy of Mid Upper Arm Circumference (MUAC) and weight-for-length Z score (WFLz) among infants aged <6 months performed by community health workers (CHWs) in Kilifi District, Kenya. METHODS: A cross-sectional repeatability study estimated inter-observer variation and accuracy of measurements initially undertaken by an expert anthropometrist, nurses and public health technicians. Then, after training, 18 CHWs (three at each of six sites) repeatedly measured MUAC, weight and length of infants aged <6 months. Intra-class correlations (ICCs) and the Pitman's statistic were calculated. RESULTS: Among CHWs, ICCs pooled across the six sites (924 infants) were 0.96 (95% CI 0.95-0.96) for MUAC and 0.71 (95% CI 0.68-0.74) for WFLz. MUAC measures by CHWs differed little from their trainers: the mean difference in MUAC was 0.65 mm (95% CI 0.023-1.07), with no significant difference in variance (P = 0.075). CONCLUSION: Mid Upper Arm Circumference is more reliably measured by CHWs than WFLz among infants aged <6 months. Further work is needed to define cut-off values based on MUAC's ability to predict mortality among younger infants.