Program derivations calculate, via step-by-step reasoning, an algorithm which implements a given specification. In practice, however, the effectiveness of calculational methods depends on the mathematical language in which they are expressed, and known techniques tend to be verbose. As a way to make calculations more approachable and easier to discover, we study the application of a graphical language of string diagrams to the process of program derivation. By having diagrams represent feasibility relations, the notion of a Galois connection, which specifies functions as computing solutions to an optimization problem, becomes very simple to represent. This enables elegant point-free specifications, as well as proofs and program derivations that are built step-by-step from equalities between diagrams. We discuss how to represent derivational techniques graphically, and propose a novel point-free induction principle suited for string diagrams. For graphical induction we need to encode additional typing information in the diagrams, which we represent using colored wires. Finally, we showcase the methodology by deriving correct-by-construction recursive algorithms for problems on natural numbers and lists.