Multisensory integration has been traditionally thought to rely on a restricted set of multisensory brain areas, and to occur automatically and pre-attentively. More recently, it has become evident that multisensory interactions can be found almost everywhere in the brain, including areas involved in attention control and areas modulated by attention. In a series of fMRI experiments, we manipulated concurrently the position of multisensory stimuli and the distribution of spatial attention. This enabled us to highlight the role of high-order fronto-parietal areas, as well as sensory-specific occipital cortex, for multisensory processing and spatial attention control. We found that specific task constraints regarding the nature of attentional deployment (endogenous vs. exogenous), the spatial relationship between stimulus position and attended location, and attentional load shape the interplay between attention and multisensory processing. We suggest that multisensory integration acts as a saliency-defining process that can interact with attentional control beyond any within-modality mechanism. We propose that an anatomically-distributed, but functionally-integrated, representation of space makes use of multisensory interactions to help attention selecting relevant spatial locations. Stimuli at the attended location undergo enhanced processing, including boosting of multisensory signals there. In this perspective, attention and multisensory integration operate in an interactive manner jointly determining the activity of a wide-spread network that includes high-order fronto-parietal regions and sensory-specific areas.