Power generation systems encompass complex designs with numerous interdependent subsystems, posing significant engineering, logistical, and management challenges. Current practices in system design and procurement often focus on component-level specifications and upfront expenditure, overlooking holistic system performance and long-term operational costs. Limited tools for design exploration and roadblocks to iterative design lead to suboptimal systems, time-consuming tasks, and operational inefficiencies. This paper introduces the Autonomous Design Optimization for Nuclear Energy (AutoDONE) framework which leverages agentic artificial intelligence (AI) and digital engineering tools to automate and optimize commercial-grade equipment selection for nuclear power plants. The framework accepts stakeholder requirements as input, identifies components that fulfill those requirements, and creates hierarchical system models for comprehensive lifecycle analysis and optimization. The paper demonstrates the capabilities of the AutoDONE framework through a test case, where the optimized set of components of a nuclear power plant system are autonomously identified and the long-term operation of the selected system is simulated. Through AI-driven intuitive automation of data processing, model generation, and decision-making, AutoDONE aims at shortening the initial design timeline by about 80% and design modification time by 50%. AutoDONE can expedite design workflows and enable engineering flexibility in nuclear power plants. The successful implementation of this framework could introduce a paradigm shift in the process of developing new nuclear facilities and upgrading existing ones, thus helping scale the deployment of power generation capabilities to meet the rising global energy demand.