Efficient information retrieval in distributed and dynamic networks remains challenging due to evolving network topologies, variable node availability, and resource constraints. In this paper, we introduce the Intelligent Reinforcement-based Mobile Agent (IRMA) framework, which utilizes Proximal Policy Optimization (PPO) to enable adaptive routing of mobile intelligent agents (MIAs). The IRMA framework was implemented and rigorously evaluated within a simulated network of 200 nodes. Comparative analyses against traditional methods—such as shortest-path, heuristic, and Deep Q-Network (DQN)-based routing—demonstrate substantial improvements. Empirical results indicate that IRMA significantly reduces latency by approximately 35%, lowers energy consumption by 25%, and enhances successful data retrieval rates by 20–30%. Statistical validation employing 95% confidence intervals and p-values below 0.05 confirms these performance enhancements. This research substantiates IRMA as a robust, scalable, and practical solution for intelligent information retrieval in dynamic network environments.