Classic predator-prey simulations done using Lotka-Volterra equations, the Wa-Tor or the Cellular Automata (CA) model, are often limited to two-species systems and lack agent memory (they are Markovian). This paper presents a novel simulation model that addresses both limitations. We extend the Wa-Tor framework to a three-species food chain (sharks, fish and plankton) and as our core innovation, integrate it with Cellular Automata with Memory (CAM). This CAM framework allows agents (sharks and fish) to use a history of recent states to make intelligent, adaptive decisions based on environmental patterns. To ensure ecological realism, the model also introduces crucial constraints, including a fish starvation mechanic and a sustainable plankton regrowth rule that emulates ecological memory. Our simulations show that the final model successfully reproduces persistent, cyclical oscillations and achieves stable long-term coexistence, as predicted by classical theory. We discuss the significant impact of agent memory on system stability and propose simulating a trophic cascade as a robust method for qualitative validation.