Artificial intelligence (AI) is reshaping predictive toxicology by enabling reproducible, automated workflows that enhance data interpretation and regulatory transparency. The O QT Assistant represents one of the first agentic AI systems capable of performing fully automated chemical hazard profiling, metabolism simulation, and read-across reasoning through the OECD QSAR Toolbox WebAPI. This presentation introduces the architecture, logic, and validation of O-QT as an open source framework designed to bridge in silico modeling with in vitro experimentation. By orchestrating large-language-model (LLM) agents to interpret molecular structures, execute profiling routines, and summarize the results in natural language, O-QT significantly reduces expert time while maintaining full traceability of each analytical step. Use cases will illustrate automated endpoint prediction and metabolism simulation across diverse chemical classes, including small organics and environmental contaminants. The system’s reasoning capabilities allow it to recommend mechanistically relevant analogues for read-across and to flag potential gaps for experimental follow-up. Beyond tool automation, the talk will discuss how multi-agent systems can improve reproducibility and foster regulatory acceptance of AI in safety assessment. The O-QT Assistant demonstrates a transparent, modular approach to integrating AI into established toxicological pipelines—transforming static QSAR workflows into dynamic, interpretable decision frameworks that advance the goals of Next-Generation Risk Assessment (NGRA).