Sepsis is defined as a dysregulated host response to infection. Decades of research have revealed many immune and metabolic mechanisms of sepsis pathogenesis, but these findings have not translated into successful therapies in the clinic. Here we work to address three major issues that have hampered this success. First, we establish a novel murine polymicrobial sepsis model that we believe addresses shortcomings of previous preclinical models: by conducting an intraperitoneal infection with the gram-negative bacterium Escherichia coli and the gram-positive bacterium Staphylococcus aureus, we combine the clinical relevance of a polymicrobial infection with the tractability of defined culturable microbes. Second, we leverage the selective serotonin reuptake inhibitor (SSRI) fluoxetine to drive immunometabolic disease tolerance during sepsis since it is widely available and affordable, has established pharmacological safety, and has demonstrated beneficial immune and metabolic effects in other contexts that translate into protective effects during sepsis. Specifically, we found that fluoxetine-mediated protection is independent of peripheral serotonin, and instead increases levels of circulating IL-10. IL-10 is necessary for protection from sepsis-induced hypertriglyceridemia and cardiac triglyceride accumulation, allowing for metabolic reprogramming of the heart. Third, we address the issue of sepsis variability – septic patients can exhibit a wide variety of outcomes, and the mechanisms that drive some individuals to survive and others to succumb are unclear. We found that in our murine polymicrobial sepsis model, cerebral cortex structure could be used to predict infection outcome – mice with thicker cortexes were more likely to survive. Together this work leverages immunometabolic mechanisms to identify a novel host-directed therapy and reveals a readily quantifiable trait that accurately trains machine learning models to predict sepsis outcome.