Objectives The purpose was to implement a fast 3D glycosaminoglycan Chemical Exchange Saturation Transfer (gagCEST) sequence at 7 T, test stability and reproducibility in cartilage in the knee in healthy volunteers, and evaluate clinical applicability in cartilage repair patients. Methods Experiments were carried out on a 7-T scanner using a volume transmit coil and a 32-channel receiver wrap-around knee coil. The 3D gagCEST measurement had an acquisition time of 7 min. Signal stability and reproducibility of the GAG effect were assessed in eight healthy volunteers. Clinical applicability of the method was demonstrated in five patients before cartilage repair surgery. Results Coefficient of variation of the gagCEST signal was 1.9%. The reproducibility of the GAG effect measurements was good in the medial condyle (ICC = 0.87) and excellent in the lateral condyle (ICC = 0.97). GAG effect measurements in healthy cartilage ranged from 2.6%-12.4% compared with 1.3%-5.1% in damaged cartilage. Difference in GAG measurement between healthy cartilage and damaged cartilage was significant ( p < 0.05). Conclusions A fast 3D gagCEST sequence was applied at 7 T for use in cartilage in the knee, acquired within a clinically feasible scan time of 7 min. We demonstrated that the method has high stability, reproducibility and clinical applicability. Key Points • gagCEST measurements are stable and reproducible • A non-invasive GAG measurement with gagCEST can be acquired in 7 min • gagCEST is able to discriminate between healthy and damaged cartilage