Restoration of ecologically important marine species and habitats is restricted by funding constraints and hindered by lack of information about trade-offs among restoration goals and the effectiveness of alternative restoration strategies. Because ecosystems provide diverse human and ecological benefits, achieving one restoration benefit may take place at the expense of other benefits. This poses challenges when attempting to allocate limited resources to optimally achieve multiple benefits, and when defining measures of restoration success. We present a restoration decision-support tool that links ecosystem prediction and human use in a flexible "optimization" framework that clarifies important restoration trade-offs, makes location-specific recommendations, predicts benefits, and quantifies the associated costs (in the form of lost opportunities). The tool is illustrated by examining restoration options related to the eastern oyster, Crassostrea virginica, which supported an historically important fishery in Chesapeake Bay and provides a range of ecosystem services such as removing seston, enhancing water clarity, and creating benthic habitat. We use an optimization approach to identify the locations where oyster restoration efforts are most likely to maximize one or more benefits such as reduction in seston, increase in light penetration, spawning stock enhancement, and harvest, subject to funding constraints and other limitations. This proof-of-concept Oyster Restoration Optimization model (ORO) incorporates predictions from three-dimensional water quality (nutrients-phytoplankton zooplankton-detritus [NPZD] with oyster filtration) and larval transport models; calculates size- and salinity-dependent growth, mortality, and fecundity of oysters; and includes economic costs of restoration efforts. Model results indicate that restoration of oysters in different regions of the Chesapeake Bay would maximize different suites of benefits due to interactions between the physical characteristics of a system and nonlinear biological processes. For example, restoration locations that maximize harvest are not the same as those that would maximize spawning stock enhancement. Although preliminary, the ORO model demonstrates that our understanding of circulation patterns, single-species population dynamics and their interactions with the ecosystem can be integrated into one quantitative framework that optimizes spending allocations and provides explicit advice along with testable predictions. The ORO model has strengths and constraints as a tool to support restoration efforts and ecosystem approaches to fisheries management.