Computing oriented plant biology is an increasingly growing field. Its applications go from the implementation of non-conventional computers to environmental exploration and monitoring, up to urban sustainable development. The interaction between plants and experimenters has to be established through the implementation of suitable interfaces. This involves, among other things, the choice of proper protocols for communication. In order to interact with the plant, the experimenter must first understand which properties of a natural stimulus make a plant reactive to the environment and, then, encode this information in an artificial signal. A way to consider the interaction between plants and machines is as an application of automata theory. This approach has already been employed to model non-biochemical reactions and processes. By extending it to machine-plant communication, I aim to identify some constitutive features and limitations of plants’ computational power. This kind of analysis is interesting for at least two reasons. First, it provides a conceptual framework for future research on human-plant interactions. Second, it can be an important building block for a theory of natural automata, aimed to identify hierarchies of computational complexity in biological phenomena.