Many plant species translocate maternally synthesized specialized metabolites to the seed to protect the developing embryo and later the germinating seedling before it initiates its own de novo synthesis. While the transport route into the seed is well established for primary metabolites, no model exists for any class of specialized metabolites that move from maternal source tissue(s) to embryo. Glucosinolate seed loading in Arabidopsis depends on plasma membrane localized exporters (USUALLY MULTIPLE AMINO ACIDS MOVE IN AND OUT TRANSPORTERs, UMAMITs) and importers (GLUCOSINOLATE TRANSPORTERs, GTRs), but the critical barriers in the seed loading process remain unknown. Here we dissect the transport route of glucosinolates from their source in the reproductive organ to the embryo by re-introducing the transporters at specific apoplastic barriers in their respective mutant backgrounds. We find that UMAMIT exporters and GTR importers form a transporter cascade that is both essential and sufficient for moving glucosinolates across at least four plasma membrane barriers along the route. We propose a model in which UMAMITs export glucosinolates out of the biosynthetic cells to the apoplast, from where GTRs import them into the phloem stream, which moves them to the unloading zone in the chalazal seed coat. From here, the UMAMITs export them out of maternal tissue and ultimately, the GTRs import them into the embryo symplasm, where the seed-specific glucosinolate profile is established by enzymatic modifications. Moreover, we propose that methylsulfinylalkyl glucosinolates are the predominant mobile form in seed loading. Elucidation of the seed loading process of glucosinolates identifies barrier-specific targets for transport engineering strategies to eliminate or over-accumulate a specialized metabolite in seeds with minimal interruption of other cellular processes.