Leptospirosis has been one of the worldwide zoonotic diseases caused by pathogenic Leptospira spp. Many molecular techniques have consecutively been developed to detect such pathogen including loop–mediated isothermal amplification method (LAMP). The objectives of this study were to evaluate the diagnostic accuracy of LAMP assay and real–time PCR using bacterial culture as a gold standard and to assess the agreement among these three tests using Cohen's kappa statistics. In total, 533 urine samples were collected from 266 beef and 267 dairy cattle reared in central region of Thailand. Sensitivity and specificity of LAMP were 96.8% (95% CI 81.5–99.8) and 97.0% (95% CI 94.9–98.2), respectively. The accuracy of LAMP (97.0%) was significantly higher than that of real–time PCR (91.9%) at 95% CI. With Cohen's kappa statistics, culture method and LAMP were substantially agreed with each other (77.4%), whereas real–time PCR only moderately agreed with culture (47.7%) and LAMP (45.3%), respectively. Consequently, LAMP was more effective than real-time PCR in detecting Leptospira spp. in the urine of cattle. Besides, LAMP had less cost and was simpler than real-time PCR. Thus, LAMP was an excellent alternative for routine surveillance of leptospirosis in cattle.