Low molecular-mass plasma proteins play a key role in health and disease. Cystatin C is an endogenous cysteine proteinase inhibitor belonging to the type 2 cystatin superfamily. The mature, active form of human cystatin C is a single non-glycosylated polypeptide chain consisting of 120 amino acid residues, with a molecular mass of 13,343-13,359 Da, and containing four characteristic disulfide-paired cysteine residues. Human cystatin C is encoded by the CST3 gene, ubiquitously expressed at moderate levels. Cystatin C monomer is present in all human body fluids; it is preferentially abundant in cerebrospinal fluid, seminal plasma, and milk. Cystatin C L68Q variant is an amyloid fibril-forming protein with a high tendency to dimerize. It forms self-aggregates with massive amyloid deposits in the brain arteries of young adults, leading to lethal cerebral hemorrhage. The main catabolic site of cystatin C is the kidney: more than 99% of the protein is cleared from the circulation by glomerular ultrafiltration and tubular reabsorption. The diagnostic value of cystatin C as a marker of kidney dysfunction has been extensively investigated in multiple clinical studies on adults, children, and in the elderly. In almost all the clinical studies, cystatin C demonstrated a better diagnostic accuracy than serum creatinine in discriminating normal from impaired kidney function, but controversial results have been obtained by comparing this protein with other indices of kidney disease, especially serum creatinine-based equations. In this review, we present and discuss most of the available data from the literature, critically reviewing conclusions and suggestions for the use of cystatin C in clinical practice. Despite the multitude of clinical data in the literature, cystatin C has not been widely used, perhaps because of a combination of factors, such as a general diffidence among clinicians, the absence of definitive cut-off values, conflicting results in clinical studies, no clear evidence on when and how to request the test, the poor commutability of results, and no accurate examination of costs and of its routine use in a stat laboratory.