This paper focuses on the consensus control issue for a class of discrete time-varying stochastic nonlinear multi-agent systems with a probability-quantization-based coding scheme. The main target is to devise a reliable control protocol, ensuring that all the agents remain confined within pre-defined ellipsoids with a certain probability despite potential coding errors and variations in feedback gain. The explicit formulation of the designed controllers and the conditions for their existence are established through recursive linear matrix inequalities. Moreover, to ensure the desired consensus performance, we address an optimization problem based on the above research. Finally, a demonstrative example is employed to show the reliability of the designed control protocol.