AugServe is an efficient inference framework that addresses the challenges of serving augmented large language models by introducing a two-stage adaptive request scheduling strategy and a dynamic token-level batching mechanism. It significantly reduces queuing latency and enhances effective throughput, leading to improved user experience in web applications.