Although cascading effects of top predators can help structure communities, their influence may vary across habitats that differentially protect prey. Therefore, to understand how and to what degree habitat complexity can affect trophic interactions in adjacent habitats, we used a combination of a broad regional-scale survey, manipulative field trials, and an outdoor mesocosm experiment to quantify predator–prey interaction strengths across four trophic levels. Within estuaries of the southeastern USA, bonnethead sharks (Sphyrna tiburo) hunt blue crabs on mudflats and adjacent oyster reefs, two habitats with vastly different aboveground structure. Using 12-h tethering trials of blue crabs we quantified habitat-dependent loss rates of 37% on reefs and 78% on mudflats. We hypothesized that the sharks’ predatory effects on blue crabs would cascade down to release a lower-level mud crab predator, which subsequently would increase juvenile oyster mortality, but that the cascade strength would be habitat-dependent. We experimentally manipulated predator combinations in split-plot mesocosms containing reef and mudflat habitats, and quantified oyster mortality. Bonnetheads exerted strong consumptive and non-consumptive effects on blue crabs, which ceased eating oysters in the sharks’ presence. However, mud crabs, regardless of shark and blue crab presence, continued to consume oysters, especially within the structural refuge of the reef where they kept oyster mortality high. Thus, bonnetheads indirectly boosted oyster survival, but only on the mudflat where mud crabs were less active. Our work demonstrates how structural differences in adjacent habitats can moderate trophic cascades, particularly when mesopredators exhibit differential use of structure and different sensitivities to top predators.