Accurate earthquake source parameters (e.g., magnitude, source location, and focal mechanism) are of key importance in seismic source studies and seismic hazard assessments. The routine workflow of source parameters estimation consists of two steps: source location inversion and focal mechanism inversion. Separate inversion of source parameters is subject to the cumulative uncertainties of both two steps inversion processes. Markov Chain Monte Carlo (MCMC), as global optimization, has been adopted in many nonlinear inversion problems to reduce cumulative errors and provide uncertainty assessment, but the application of MCMC is strongly subject to prior information. In this study, we present a new Python package MCMTpy. MCMTpy exploits the Cut‐And‐Paste (CAP) algorithm and Bayesian inference, using Markov Chain to implement the source location inversion and focal mechanism inversion in one inversion workflow. The new approach can effectively reduce the prior model dependence, and is closely integrated into the current seismological programming ecosystem. To demonstrate the effectiveness of the new package, we applied the MCMTpy to the 2021 6.4 Yangbi earthquake, Yunnan, China, and 2008 5.2 Mt. Carmel Earthquake, Illinois. A comparison between our results and other catalogs (e.g., Global Centroid Moment Tensor and U.S. Geological Survey W‐phase) solutions illustrates that both double‐couple and moment tensor solutions can be reliably recovered. The robustness and limitations of our approach are demonstrated by an experiment with 30 different initial models and an experiment with the grid‐search method.