Cloud presence in optical satellite imagery poses significant challenges for various earth observation tasks. During training, this work utilized a mamba-based conditional GAN for optical satellite cloud removal by leveraging paired Sentinel-2 images (one containing clouds and the other cloud-free). State space model based mamba block helps in better pixel relation extraction. The proposed model exploits spatial and spectral correlations across multiple bands to reconstruct cloud-covered areas accurately. By utilizing cloud-unaffected spectral bands and optimizing loss functions for detail preservation, our approach achieves high-quality cloud-free reconstructions, outperforming traditional cloud-removal techniques. This method enhances optical data’s temporal availability and quality, enabling more consistent and accurate analysis in various remote sensing applications. Our curated dataset can be shared for future research.