Foraging decisions by rodents are key for the long-term maintenance of oak populations in which avian seed dispersers are absent or inefficient. Decisions are determined by the environmental setting in which acorn-rodent encounters occur. In particular, seed value, competition and predation risks have been found to modify rodent foraging decisions in forest and human-modified habitats. Nonetheless, there is little information about their joint effects on rodent behavior, and hence, local acorn dispersal (or predation). In this work, we manipulate and model the mouse-oak interaction in a Spanish dehesa, an anthropogenic savanna system in which nearby areas can show contrasting levels of ungulate densities and antipredatory cover. First, we conducted a large-scale cafeteria field experiment, where we modified ungulate presence and predation risk, and followed mouse foraging decisions under contrasting levels of moonlight and acorn availability. Then, we estimated the net effects of competition and risk by means of a transition probability model that simulated mouse foraging decisions. Our results show that mice are able to adapt their foraging decisions to the environmental context, affecting initial fates of handled acorns. Under high predation risks mice foraged opportunistically carrying away large and small seeds, whereas under safe conditions large acorns tended to be predated in situ. In addition, in the presence of ungulates lack of antipredatory cover around trees reduced mice activity outside tree canopies, and hence, large acorns had a higher probability of survival. Overall, our results point out that inter-specific interactions preventing efficient foraging by scatter-hoarders can reduce acorn predation. This suggests that the maintenance of the full set of seed consumers as well as top predators in dehesas may be key for promoting local dispersal.