How situated, embodied agents may achieve goals using knowledge is a central challenge for a neural theory of natural and artificial intelligence. A new paper by Jan Tekülve and Gregor Schöner borrows terms from Searle’s analysis of intentionality to structure a neural dynamic architecture that illustrates a possible route to such a theory. Intentional states are instantiated by neural activation patterns that are stabilized by neural interaction. Their dynamic instabilities enable the autonomous generation of sequences of intentional states. In a toy demonstration, an intentional agent is grounded in perception and action and generates memories and action plans. It autonomously learns beliefs that it activates to achieve desired outcomes.