In any engineering alloy, a propagating crack will encounter a multitude of microstructural features like point defects, dislocations, precipitates, grain and phase boundaries. All of these interactions will influence the crack and contribute to energy dissipation. Understanding the elementary interaction mechanisms between cracks and these defects is therefore of fundamental importance for the development of microstructure-sensitive micromechanical failure models. Such models in turn are key elements for any physics-based description of the process-microstructure-damage linkages and for the design of damage-tolerant materials. Understanding the unit processes in crack–obstacle interactions can, however, only be the first step. Predictive failure models would need to also include the statistical distribution of defects and their interactions with each other in the stress field of the crack. Here, we show an example of how atomistic simulations of crack–microstructure interactions can be used to inform a mesoscale model, which in turn can be compared to dedicated micromechanical fracture experiments. Atomistic simulations on various bcc metals show that point defects not only lead to solid- solution-like hardening but also directly affect crack propagation. The interaction of dislocations with cracks can lead to cross-slip processes, crack-front reorientation and kinking as well as to the nucleation of new dislocations. Similarly, the interaction of propagating cracks with localized obstacles like voids can also lead to the nucleation of dislocations, crack deflection and crack arrest. Such processes are included in a newly developed code that couples discrete dislocations dynamics (DDD) with the extended finite element method (XFEM) for fracture simulations. This mesoscale model is then used to qualitatively study the effect of varying the dislocation and void density on the fracture toughness as determined by micromechanical tests on notched bending beams.