The rapid evolution of wireless networks presents unprecedented challenges in managing diverse wireless tasks. These challenges underscore the need for AI-native solutions in next-generation networks. In this article, we propose WirelessAgent, a novel framework that harnesses large language models (LLMs) to create autonomous AI agents for diverse wireless network tasks. We first define a general framework for WirelessAgent, supported by key components and principles in AI agents. Then, we introduce a basic usage to implement the WirelessAgent based on agentic workflows and the LangGraph architecture. We demonstrate the effectiveness of WirelessAgent through a comprehensive case study on the network slicing task. Our numerical results show that WirelessAgent achieves 44.4% higher bandwidth utilization than the Prompt-based method, while performing only 4.3% below the Rule-based optimality. Notably, WirelessAgent delivers near-optimal network throughput across diverse network scenarios. These underscore the framework's potential for intelligent and autonomous control in next-generation networks. The code is available at https://github.com/jwentong/WirelessAgent_R1.