CellRank 2 is a versatile and scalable framework for studying cellular fate using multiview single-cell data. It integrates multiple data modalities, including RNA velocity, experimental time points, and metabolic labeling, to infer cell-state dynamics and fate probabilities. The framework allows combining transitions across and within experimental time points, enabling the identification of terminal states and fate probabilities. It also estimates cell-specific transcription and degradation rates from metabolic-labeling data, which is applied to intestinal organoids to delineate differentiation trajectories. CellRank 2 decomposes trajectory inference into modality-specific modeling of cell transitions and modality-agnostic trajectory inference. It uses a modular design with kernels for computing cell-cell transition matrices and estimators for analyzing these matrices to identify initial and terminal states, fate probabilities, and lineage-correlated genes. The framework is robust, scalable, and applicable to a wide range of data modalities, including pseudotime, developmental potentials, and experimental time points. The PseudotimeKernel improves upon RNA velocity by computing pseudotime-informed transition probabilities, allowing accurate recovery of terminal states and fate probabilities in hematopoiesis. The CytoTRACEKernel enables the inference of developmental potentials and terminal states in endodermal development. The RealTimeKernel combines inter- and intra-time-point transitions to provide a more granular mapping of cell fate, improving the identification of terminal states and initial states. CellRank 2 also estimates kinetic rates from metabolic-labeling data, enabling the analysis of gene regulatory strategies underlying cellular state changes. It outperforms existing methods in terms of accuracy and scalability, particularly in identifying terminal states and lineage-correlated genes. The framework supports the integration of multiple data modalities, allowing for a more comprehensive understanding of cellular dynamics and fate decisions. Overall, CellRank 2 provides a unified and flexible approach for analyzing complex single-cell data across various biological systems.CellRank 2 is a versatile and scalable framework for studying cellular fate using multiview single-cell data. It integrates multiple data modalities, including RNA velocity, experimental time points, and metabolic labeling, to infer cell-state dynamics and fate probabilities. The framework allows combining transitions across and within experimental time points, enabling the identification of terminal states and fate probabilities. It also estimates cell-specific transcription and degradation rates from metabolic-labeling data, which is applied to intestinal organoids to delineate differentiation trajectories. CellRank 2 decomposes trajectory inference into modality-specific modeling of cell transitions and modality-agnostic trajectory inference. It uses a modular design with kernels for computing cell-cell transition matrices and estimators for analyzing these matrices to identify initial and terminal states, fate probabilities, and lineage-correlated genes. The framework is robust, scalable, and applicable to a wide range of data modalities, including pseudotime, developmental potentials, and experimental time points. The PseudotimeKernel improves upon RNA velocity by computing pseudotime-informed transition probabilities, allowing accurate recovery of terminal states and fate probabilities in hematopoiesis. The CytoTRACEKernel enables the inference of developmental potentials and terminal states in endodermal development. The RealTimeKernel combines inter- and intra-time-point transitions to provide a more granular mapping of cell fate, improving the identification of terminal states and initial states. CellRank 2 also estimates kinetic rates from metabolic-labeling data, enabling the analysis of gene regulatory strategies underlying cellular state changes. It outperforms existing methods in terms of accuracy and scalability, particularly in identifying terminal states and lineage-correlated genes. The framework supports the integration of multiple data modalities, allowing for a more comprehensive understanding of cellular dynamics and fate decisions. Overall, CellRank 2 provides a unified and flexible approach for analyzing complex single-cell data across various biological systems.