We developed agent-based models patterned after the equation-based models developed by Schimel and Weintraub (2003) to explore the influence of microbial-derived extracellular enzymes on carbon (C) dynamics. The models featured spatial arrangements of detritus as either randomly-spaced particles (rain) or as root-like structures (root), detritus input intervals (continuous vs. pulsed) and rates (0-5000 units in 500 unit intervals), trophic structures (presence or absence of predators preying on microbes), and extracellular enzymes with different half-lives (1, 10, 100, and 1000 time steps). We studied how these features affected C dynamics and model persistence (no extinctions). Models without predators were more likely to persist than those with predators, and their C dynamics could be explained with energetics-based arguments. When predators were present, two of the four model configurations – root-continuous and rain-pulsed –were more likely to persistent. The root-continuous models were more likely to persist at lower detritus input rates (500-3500 units), while the rain-pulsed models were more likely to persist at intermediate detritus input rates (2000-3500 units). For both these model configurations, shorter extracellular enzymes half-lives increased the likelihood of persistence. Consistent with the results of Schimel and Weintraub (2003), C dynamics were governed by extracellular enzyme production activity and loss. Our results demonstrated that extracellular enzymes control of C dynamics depends on the spatial arrangement of resources, the input rate and input intervals of detritus and trophic structure.