ABSTRACT Understanding how extinct animals moved is a central goal in paleobiology, yet interpreting locomotor function from anatomy alone is complicated by convergent and divergent morphologies. One promising approach is the construction of morphofunctional spaces (MFSs), which integrate multiple biomechanical indices and comparative statistics to refine functional inference. This study investigates forelimb adaptations for digging in Caraguatypotherium munozi (Notoungulata, Mesotheriidae), a mid‐sized Miocene notoungulate lacking extant analogs. We developed an MFS based on osteological measurements and mechanical advantage (MA) models at the elbow and wrist. These were derived from fossil material and comparative data across 38 extant mammal species representing 21 families and 5 locomotor habits—terrestrial, fossorial (digger), climbing, swimming, and flying—as well as 5 mesotheriid specimens, including the holotype of C. munozi . Multivariate and inferential statistical analyses were used to identify functional patterns and evaluate locomotor hypotheses. Results show that C. munozi occupies an intermediate position in MFS, adjacent to but outside the core regions of extant fossorial, climbing, and terrestrial mammals. It exhibits the highest wrist flexor MA (31.4%) in the data set, 9%–13% above the range of living scratch‐diggers, and a low elbow extensor MA (~19%), below the fossorial mean (~31.6%). Notably, incorporating manus length (MTCIII‐L) into the elbow model lowers MA further but places C. munozi within the statistical range of extant diggers, suggesting partial mechanical similarity. This biomechanical pattern supports a wrist‐dominant excavation strategy, reflecting a distinct mechanical pathway that enabled C. munozi to perform the motor gesture of scratch‐digging through enhanced distal force generation, rather than relying on proximal joint leverage as in extant fossorial mammals. The integration of MA with osteological indices within a multivariate framework provides novel insights into extinct mammalian locomotion and underscores the utility of MFS models for reconstructing context‐dependent motor capabilities and locomotor habits.