Earthquake activities in the Andaman - Nicobar region are generated mainly by several active faults and an active subduction zone that can activate multiple hypocenter clusters. Therefore, studying the seismotectonic process from source mechanism analysis using moment tensor inversion is essential. In this research, we applied a stochastic Bayesian inversion using Grond, a framework tool for moment tensor (MT) inversion, to characterize the earthquake mechanisms and the most appropriate centroid locations. Procedurally, several of the largest earthquakes with a moment magnitude Mw > 5.5 from 2007 to 2023 were analyzed using the regional and teleseismic networks up to 2000 km from the epicenter based on 1 Hz green function model and azimuthal gap coverage. We applied 10,000-30,000 iterations of inversions with 300 bootstrap chains for a stable result. The low-frequency bandpass with a range of 0.007 - 0.03 Hz was chosen for the inversion process. There were 17 solutions for earthquake mechanisms with good misfit values of < 0.65. Based on the division of the 5 clusters with the highest seismicity, the selected earthquakes’ source mechanism is dominated by strike-slip and oblique faulting. The results were compared with existing data catalogs showing differences in hypocentre and fault orientation direction, particularly for strike-slip earthquakes in the shallow crust and mantle zones.