A new caries assessment instrument, the Caries Assessment Spectrum and Treatment (CAST), was developed. It covers carious lesion progression from no lesion, sealants and restorations to lesions in enamel and dentine, advanced stages in pulpal and tooth-surrounding tissues, and tooth loss owing to dental caries, in nine codes. The objective of this study was to determine the reproducibility of the CAST instrument in primary and permanent dentitions, using three age groups. Two epidemiological surveys were conducted in Brazil, covering three age groups: 2–6-year-old and 6–9-year-old children and 19–30-year-old adults. Four trained and calibrated examiners performed the examinations. Reproducibility was calculated for intra- and inter-examiner at surface and tooth levels and expressed as unweighted kappa-coefficient value (κ) and percentage of agreement (Po) for CAST codes (0–7) and for the categories healthy (0–2) versus diseased (3–7), and non-cavitated (0–3) versus cavitated (4–7) teeth. Using CAST codes (0–7) for the 2–6-year-old age group in primary dentitions, inter-examiner consistency was κ = 0.74 and Po was 98.3%. In the 6–9-year-old age group in primary dentitions, inter-examiner consistency ranged from κ = 0.68 to κ = 0.86 and Po was ≥93.7%. In the 19–30-year-old age group inter-examiner consistency was κ = 0.87 and Po was 94.1%. The reproducibility of the CAST instrument for use in the primary dentition of 2–6-year olds and of 6–9-year olds was ‘substantial’ to ‘almost perfect’. The reproducibility for its use in the permanent dentition of 19–30-year olds was ‘almost perfect’. The CAST instrument can reliably be applied in epidemiological studies covering these ages.