**Introduction** Automatically extracting clinical data from oncology notes is a complex task requiring context and specialized terminology understanding. However, manually executing this task is costly and impractical. **New approach** A team of researchers has developed a new method for automatically extracting clinical data from oncology notes based on large language models (LLMs). This approach uses agential agents to decompose extraction tasks into modular and adaptive tasks. Additionally, the approach includes context-sensitive retrieval and iterative synthesis capabilities. **Results and applications** The new method was evaluated on a dataset of over 400,000 clinical notes and PDF reports spanning over 2,250 cancer patients. The results indicate that the approach can achieve an average F1-score of 0.93 with specific clinical variables exceeding 0.85 and critical variables (e.g., biomarkers and medications) surpassing 0.95. Furthermore, integrating the approach into a data curation workflow resulted in a 94% direct manual approval rate. **Conclusion** The new method for automatically extracting clinical data from oncology notes represents a significant turning point in clinical analysis. With its scalable and precise extraction capabilities, this approach can revolutionize how clinical data is managed and supports diagnostic and therapeutic decision-making.