Pancreas segmentation is vital for improving the accuracy and reliability of pediatric diagnostic imaging. No reliable open-source algorithms are available for segmenting the pediatric pancreas from MRI scans. Existing research focuses on CT scans, limiting their relevance to pediatric populations. We represent the first validated AI algorithm for 3D pancreas MRI segmentation in children. PanSegNet achieved mean Dice Similarity Coefficient (DSC) scores of 88% in healthy controls, 81% in acute pancreatitis (AP), and 80% in chronic pancreatitis (CP).Pancreas MRI T2W scans were obtained from 42 children with AP or CP and 42 healthy children into the study. PanSegNet achieved mean DSC scores of 88% for healthy controls, 81% for AP, and 80% for CP, showing high overlap with manual segmentations. Inter-observer kappa was 0.86 for controls and 0.82 for pancreatitis; intra-observer agreements were 0.88 and 0.81. Strong correlations were observed between automated and manual volumes (R² = 0.85 for controls, 0.77 for diseased). PanSegNet was shown to effectively segment the pancreas in children's MRI scans with high accuracy. This open-source tool and its annotated dataset fill a significant gap and enhance clinical outcomes by improving diagnostic precision for pancreatic diseases in children.