Bayesian modeling requires an iterative workflow of model exploration, validation through diagnostics, and evidence-based refinement. While principled, this process demands substantial manual effort and expertise. We present a multi-agent system that automates the complete Bayesian workflow while maintaining interpretability. Given a dataset, specialized agents coordinate to explore model alternatives, implement probabilistic programs, run inference, interpret diagnostics, and iterate based on validation results. The system produces both validated models and exploration traces documenting all modeling decisions. In experiments on PosteriorDB datasets, our system achieves competitive predictive performance while generating complete audit trails of the modeling process.