There were never real Kantians, unless we count more or less faithful popularizations of Kant's ideas like Herbart's, or sympathetic commentary like Strawson's. Helmholtz, Grassman, Poincare and other scientists were mostly interested in Kant's theory of visual perception, psychologically and physiologically reinterpreted. Perhaps the most minimal definition is given in Wikipedia without citation: ""Neo-Kantian" can also be used as a general term to designate anyone who adopts Kantian views in a partial or limited way". By this definition however Hegel and Quine are neo-Kantians, as are most contemporary philosophers.
Historically however two major philosophical schools self-described as neo-Kantians, Marburg (Cohen, Natorp, early Cassirer) and Southwest (Windelband, Rickert, Lask), and they shared with Kant two main themes. First, epistemology subsumes ontology, our beliefs are not representations of any mind-independent "objects". The very concept of such a representation is meaningless since our mental faculties can not "represent" anything independent of themselves, they only deal with appearances. Second, the object of knowledge is constituted by our mental faculties based on some a priori structures. Where neo-Kantians diverged from Kant is in rejecting a priori forms of perception, space and time, they were left therefore with undifferentiated manifold of sensation and logical forms of judgement to do the constitution.
This has far reaching consequences because it destroys the unity of Kant's architectonic of pure and practical reason, productive ability of imagination can no longer do the work for Kant's noumenal metaphysics of practical reason and judgement. Marburgers gave priority to pure reason, and focused on scientific knowledge, Southwesterners to practical reason, values and transcendental psychology over logic. Both retained priority of rational discourse over religious, artistic, etc. forms of expression, but in mild forms. Rickert, Lask and Natorp paved the way for phenomenology (personally influencing Husserl and Heidegger), Cassirer in 1920s moved beyond Marburg by incorporating Hegel's historical dialectic and elements of Lebensphilosophie, and advocating plurality of expressive forms.
By historical measure Foucault and Kuhn are frankly a stretch. The closest thing to a contemporary neo-Kantian would be Friedman, who seeks to enrich Marburg with insights from Carnap, Quine and Kuhn, and takes Cassirer as inspiration. In particular, he relativizes a priori forms and analytic/synthetic distinction to lasting but historically revisable notions that help bridge the "incommensurability" of paradigms and mitigate cultural relativism.
In Parting of the Ways Friedman gives a sympathetic account of historical neo-Kantianism and its precipitation of the analytic/continental divide despite Cassirer's best efforts.
