TorchCP provides researchers with conformal prediction tools built on PyTorch for generating statistically rigorous prediction sets with guaranteed coverage on unseen data. It supports classification, regression, and graph-structured data tasks, and includes adapters for uncertainty quantification in large language models.