Astrocytes represent a major glial cell population that regulates synaptic transmission, neural plasticity, and homeostasis through complex spatiotemporal interactions with neuronal networks. This paper presents a novel Multi-Agent Reinforcement Learning (MARL) framework for modelling astrocytic behaviour within neural microenvironments. In this simulation, individual astrocytes are modeled as autonomous learning agents that independently regulate glutamate dynamics while coordinating with neighbouring astrocytic domains. The model is evaluated by comparing glutamate release rates against standard reinforcement learning approaches. including independent proximal policy optimisation (IPPO), QMIX, and Value Decomposition Networks(VDN). Experimental results indicated that the adaptive IPPO based method achieves around 1% deviation from reported values. The proposed computational model addresses the dual role of astrocytes in maintaining glutamate homeostasis under physiological conditions while potentially contributing to neuronal damage in disease conditions. This study demonstrates the potential of the framework for applications in organ-on-chip platforms and therapeutic intervention research.