This paper proposes MTS-CGAN, a Conditional Generative Adversarial Network for Multivariate Time Series, with both the generator and discriminator networks based on Transformers. MTS-CGAN leverages encoded context for conditional generation, allowing the generation of complex time series in a single pass and outperforming existing models, particularly in mixed distribution contexts. We evaluate MTS-CGAN using quantitative and qualitative metrics across various multivariate time series datasets. Additionally, we introduce an innovative adaptation of the Frechet Inception Distance (FID) for time series, providing a robust measure of the quality of generated data. This research highlights the potential of MTS-CGAN in generating high-quality multivariate time series.