PROBCONS is a novel tool for protein multiple sequence alignment that introduces the concept of *probabilistic consistency*. This scoring function is based on a pair hidden Markov model (HMM) and aims to improve alignment accuracy by incorporating information from multiple sequences. The algorithm first computes posterior probability matrices for pairwise alignments, then uses these matrices to compute expected accuracies. The probabilistic consistency transformation is applied to adjust match quality scores based on information from a third sequence. This process is repeated iteratively to enhance alignment reliability. PROBCONS is evaluated on several benchmark datasets, including BAliBASE, SABmark, and PREFAB, and shows statistically significant improvements in accuracy over other leading methods while maintaining practical running times. The tool is publicly available under the GNU Public License.PROBCONS is a novel tool for protein multiple sequence alignment that introduces the concept of *probabilistic consistency*. This scoring function is based on a pair hidden Markov model (HMM) and aims to improve alignment accuracy by incorporating information from multiple sequences. The algorithm first computes posterior probability matrices for pairwise alignments, then uses these matrices to compute expected accuracies. The probabilistic consistency transformation is applied to adjust match quality scores based on information from a third sequence. This process is repeated iteratively to enhance alignment reliability. PROBCONS is evaluated on several benchmark datasets, including BAliBASE, SABmark, and PREFAB, and shows statistically significant improvements in accuracy over other leading methods while maintaining practical running times. The tool is publicly available under the GNU Public License.