One paper accepted to CoRL 2022! We show how Transformers can combine state history, multiple camera views, and natural language instructions to perform a variety of manipulation tasks on the RLBench benchmark. ArXiv