A method based on surface enhanced laser desorption and ionization time of flight mass spectrometry (SELDI-TOF MS) was developed for the rapid identification of Klebsiella pneumoniae by directly applying bacterial colonies without further protein extraction. A total of 40 K. pneumoniae and 114 other related microorganisms isolated clinically were analyzed by SELDI-TOF MS. An identification model for K. pneumoniae was established by artificial neural networks (ANNs) with classification accuracy of 100%. The model was blindly tested with 43 K. pneumoniae and 53 control bacteria again. The results showed that the model was successful with accuracy of 96.9%, sensitivity of 100% and specificity of 94.3%. This strategy is potential for rapid identification of K. pneumoniae.