CLAM is a weakly-supervised, data-efficient, and interpretable deep learning method for whole slide image (WSI) analysis. It addresses key challenges in computational pathology, including weak supervision, data efficiency, multi-class subtyping, adaptability, and interpretability. CLAM uses attention-based learning to identify high-value sub-regions in WSIs for accurate classification, while also utilizing instance-level clustering to refine the feature space. It requires only slide-level labels and can handle multi-class subtyping tasks. CLAM is adaptable to various data sources, including cell phone microscopy images, biopsy slides, and different tissue content. It produces interpretable heatmaps that highlight regions important for diagnosis without requiring pixel-level annotations. CLAM has been tested on three tasks: renal cell carcinoma subtyping, non-small cell lung cancer subtyping, and lymph node metastasis detection. It achieves high performance across these tasks with minimal training data. CLAM is also adaptable to independent test cohorts and can generalize to new data sources. It is publicly available as an easy-to-use Python package and can be visualized through an interactive demo. CLAM is a flexible and general-purpose method that can be applied to various computational pathology tasks in clinical and research settings.CLAM is a weakly-supervised, data-efficient, and interpretable deep learning method for whole slide image (WSI) analysis. It addresses key challenges in computational pathology, including weak supervision, data efficiency, multi-class subtyping, adaptability, and interpretability. CLAM uses attention-based learning to identify high-value sub-regions in WSIs for accurate classification, while also utilizing instance-level clustering to refine the feature space. It requires only slide-level labels and can handle multi-class subtyping tasks. CLAM is adaptable to various data sources, including cell phone microscopy images, biopsy slides, and different tissue content. It produces interpretable heatmaps that highlight regions important for diagnosis without requiring pixel-level annotations. CLAM has been tested on three tasks: renal cell carcinoma subtyping, non-small cell lung cancer subtyping, and lymph node metastasis detection. It achieves high performance across these tasks with minimal training data. CLAM is also adaptable to independent test cohorts and can generalize to new data sources. It is publicly available as an easy-to-use Python package and can be visualized through an interactive demo. CLAM is a flexible and general-purpose method that can be applied to various computational pathology tasks in clinical and research settings.