Objective To design a simple magnetic resonance (MR)–based assessment system for quantification of osteochondral defect severity prior to cartilage repair surgery at the knee. Design The new scoring tool was supposed to include 3 different parameters: (1) cartilage defect size, (2) depth/morphology of the cartilage defect, and (3) subchondral bone quality, resulting in a specific 3-digit code. A clearly defined numeric score was developed, resulting in a final score of 0 to 100. Defect severity grades I through IV were defined. For intra- and interobserver agreement, defects were assessed by 2 independent readers on preoperative knee MR images of n = 44 subjects who subsequently received cartilage repair surgery. For statistical analyses, mean values ± standard deviation (SD), interclass correlation coefficients (ICC), and linear weighted kappa values were calculated. Results The mean total Area Measurement And DEpth & Underlying Structures (AMADEUS) score was 48 ± 24, (range, 0-85). The mean defect size was 2.8 ± 2.6 cm². There were 36 of 44 full-thickness defects. The subchondral bone showed defects in 21 of 44 cases. Kappa values for intraobserver reliability ranged between 0.82 and 0.94. Kappa values for interobserver reliability ranged between 0.38 and 0.85. Kappa values for AMADEUS grade were 0.75 and 0.67 for intra- and interobserver agreement, respectively. ICC scores for the AMADEUS total score were 0.97 and 0.96 for intra- and interobserver agreement, respectively. Conclusions The AMADEUS score and classification system allows reliable severity encoding, scoring and grading of osteochondral defects on knee MR images, which is easily clinically applicable in daily practice.