Arthroplasty surgery can be challenging because of limited exposure of the bones. However, in shoulder arthroplasty for example, a good positioning of the glenoid component is essential to mitigate risks of revision surgeries. To improve the procedure's outcomes, mixed reality can be used as a guidance system relying on a fundamental registration step between the patient's bone and its corresponding 3D model. We present a complete registration workflow for shoulder arthroplasty using Hololens 2 Head Mounted Display. We rely on acquisitions made thanks to our marker-based tracking system, an improved Iterative Closest Point algorithm and verification steps. Our accuracy targets are 1.5 ± 1.5mm for the glenoid guidewire entry point on both antero-posterior and supero-inferior axes, and 1.5 ± 1[Formula: see text] for inclination and version. The overall process must last less than 5min. We have evaluated our process on a cohort of 13 3D printed glenoid bones of all types, showing an average accuracy of 0.84 ± 0.58mm on the antero-posterior axis and 0.49 ± 0.41mm on the supero-inferior one for the entry point. As for inclination and version, we have 0.89±0.6[Formula: see text] and 0.97±0.8[Formula: see text], respectively. The mean process time is about 1min 24s. We have developed a complete, embedded registration workflow and a verification protocol to evaluate our accuracy. Our results are promising for the improvement of the glenoid guidewire placement. Moreover, everything is performed in the field of view of the surgeon, which allows them to fully concentrate on the surgical site.