Unreinforced masonry (URM) buildings that are characterized by brittle, heterogeneous, and anisotropic responses have complex nonlinear dynamic behavior and high vulnerability under seismic loading. While masonry confined by tie‐columns and lintel beams can achieve better ductility and stability, the wall openings induce stress concentration around the corners and diminish the strength. Furthermore, the interacting in‐plane and out‐of‐plane responses of masonry walls could significantly affect the post‐cracking behavior and increase the collapse risk. To attain reliable seismic design of confined masonry buildings with openings, the collapse fragility under bi‐directional seismic loading shall be addressed and thoroughly investigated in this study. Advanced discrete finite element models (DFEM) with coupled damage‐plasticity traction‐separation law, which can simulate complex nonlinear dynamic behavior such as disintegration and progressive collapse, were developed for the masonry structures and validated against experimental results. Four one‐storey residential masonry buildings: (1) unconfined (UCM), (2) confined by tie‐columns at four building corners (CM‐I), (3) confined by toothed tie‐columns (CM‐II), and (4) confined by toothed tie‐columns and lintel beams (CM‐III) were considered. A series of incremental dynamic analyses (IDA) with 15 bi‐directional ground motions and fragility analyses were performed to investigate the dynamic characteristics, base shear, hysteresis behavior, failure mechanism, collapse probability, and exceedance probability for different drift limits were studied in‐depth. CM‐II with toothed tie‐columns had the lowest collapse probability and fragility at different drift limit states. The introduction of lintel beams to CM‐III lead to an increase of the storey‐drift‐based fragility compared to CM‐II, as a result of the increased maximum base shear of CM‐III with higher post‐damage stiffness. The lintel beams also inflicted short‐column effects on the tie‐columns. Nevertheless, the lintel beams could delay the out‐of‐plane collapse of the masonry walls by shortening the effective wall arching length. In some cases, the CM‐III could achieve better performances than CM‐II even in terms of storey drifts. The response variations of CM‐III to the changing of the non‐intensity related ground motion parameters are also lower than CM‐II.