Management strategy evaluation (MSE) provides a simulation framework to test the performance of living marine resource management. MSE has now been adopted broadly for use in single-species fishery management, often using a relatively simple “operating model” that projects population dynamics of one species forward in time. However, many challenges in ecosystem-based management involve tradeoffs between multiple species and interactions of multiple stressors. Here we use complex operating models, multi-species ecosystem models of the California Current and Nordic and Barents Seas, to test threshold harvest control rules that explicitly address the linkage between predators and prey, and between the forage needs of predators and fisheries. Specifically, within Atlantis ecosystem models we focus on how forage (zooplankton) availability affects the performance of harvest rules for target fish, and how these harvest rules for fish can account for environmentally-driven fluctuations in zooplankton. Our investigation led to three main results. First, consistent with studies based on single-species operating models, we found that compared to constant F = FMSY policies, threshold rules led to higher target stock biomass for Pacific hake (Merluccius productus) in the California Current and mackerel (Scomber scombrus) in the Nordic and Barents Seas. Performance in terms of catch of these species varied depending partly on the biomass and recovery trajectory for the simulated stock. Secondly, the multi-species operating models and the harvest control rules that linked fishing mortality rates to prey biomass (zooplankton) led to increased catch variability; this stemmed directly from the harvest rule that frequently adjusted Pacific hake or mackerel fishing rates in response to zooplankton, which are quite variable in these two ecosystems. Thirdly, tests suggested that threshold rules that increased fishing when productivity (zooplankton) declined had the potential for strong ecosystem effects on other species. These effects were most apparent in the Nordic and Barents Seas simulations. The tests of harvest control rules here do not include uncertainty in monitoring of fish and zooplankton, nor do they include uncertainty in stock assessment and implementation; these would be required for full MSE. Additionally, we intentionally chose target fish with strong mechanistic links to particular zooplankton groups, with the simplifying assumption that zooplankton biomass followed a forced time series. Further developing and testing of ecosystem-level considerations can be achieved with end-to-end ecosystem models, such as the Atlantis models applied here, which have the added benefit of tracking the follow-on effects of the harvest control rule on the broader ecosystem.