A 50 × 50 × 10 mm3 monolithic gadolinium aluminum gallium garnet (Gd3Al2Ga3O12; GAGG):Ce crystal coupled to a 8 × 8 silicon photomultiplier (SiPM) array was developed; it showed very good system uniformity and a high energy resolution of 7.4% at 662 keV. By using a convolutional neural network-based positioning algorithm and a fan-beam calibration method, the detector achieved a position resolution of ∼1.4 mm and a depth of interaction resolution of ∼2 mm. Based on this high-performance monolithic detector, we developed a coded aperture gamma camera. A 1-mCi Cs-137 source centered at a 2-m distance from the mask could be reconstructed with a signal-to-noise ratio of 6.5 in 1 s. Furthermore, the imaging ability of a low-energy Am-241 source and a low-activity Cs-137 source when the background-to-signal ratio was approximately 1:1 and a double low-activity source (Cs-137 and Na-22) was demonstrated. It is shown that the monolithic-crystal-based coded aperture gamma camera can achieve high performance and has a large potential for further improvement.