Abstract Background With advances in next generation sequencing technology and analysis methods, single nucleotide variants (SNVs) and indels can be detected with high sensitivity and specificity in exome sequencing data. Recent studies have demonstrated the ability to detect disease-causing copy number variants (CNVs) in exome sequencing data. However, exonic CNV prediction programs have shown high false positive CNV counts, which is the major limiting factor for the applicability of these programs in clinical studies. Results We have developed a tool (cnvScan) to improve the clinical utility of computational CNV prediction in exome data. cnvScan can accept input from any CNV prediction program. cnvScan consists of two steps: CNV screening and CNV annotation. CNV screening evaluates CNV prediction using quality scores and refines this using an in-house CNV database, which greatly reduces the false positive rate. The annotation step provides functionally and clinically relevant information using multiple source datasets. We assessed the performance of cnvScan on CNV predictions from five different prediction programs using 64 exomes from Primary Immunodeficiency (PIDD) patients, and identified PIDD-causing CNVs in three individuals from two different families. Conclusions In summary, cnvScan reduces the time and effort required to detect disease-causing CNVs by reducing the false positive count and providing annotation. This improves the clinical utility of CNV detection in exome data.