This article reconsiders the theoretical role of the genetic code. By drawing on published and unpublished sources from the 1950s, I analyse how the code metaphor was actually employed by the scientists who first promoted its use. The analysis shows that the term ‘code’ picked out mechanism sketches, consisting of more or less detailed descriptions of ordinary molecular components, processes, and structural properties of the mechanism of protein synthesis. The sketches provided how-possibly explanations for the ordering of amino acids by nucleic acids (the ‘coding problem’). I argue that employing the code metaphor was justified in virtue of its descriptive-denotational and explanatory roles, and because it highlighted a similarity with conventional codes that was particularly salient at the time.