Abstract: Seismic fragility function is an important tool for measuring the probabilistic seismic performance of structures, and its accuracy affects earthquake risk assessment and disaster reduction strategies. Bayesian method can improve the accuracy of fragility functions by incorporating observed data such as post-earthquake damage investigation or large-scale experiments. However, the current numerical methods for posterior estimation of the fragility function parameters lack specific steps and clear parameter settings. To this end, Bayesian updating method for seismic fragility functions based on Markov Chain Monte Carlo is systematically presented. Firstly, seismic fragility function and its parameters were introduced; Secondly, Bayesian updating method for fragility functions, including the prior distribution, likelihood function, and posterior distribution were elaborated. Thirdly, the specific steps and sampling strategies for calculating the Bayesian posterior estimation of the fragility function parameters using Markov Chain Monte Carlo simulation were given. Finally, the implementation process with this method to update the fragility function of a multi-story steel frame structure based on the observed data from a full-scale shaking table test were presented, including the analyses of the updating effect of the fragility function parameters. The results show that calculating the Bayesian posterior estimation with Markov Chain Monte Carlo to correct the two parameters of the fragility function is effective, and the updating effect of the median value is significant, with the highest reaching 35% in this example. However, the updating effect of the logarithmic standard deviation is not obvious. The more evidence in the observed data that exceed a certain limit state, the higher the confidence level of the posterior estimated fragility function for that state in Bayesian updating. This research can provide a reference for the assessment of structural seismic performance by integrating multiple sources of data.