Dynamic assessment (DA) derives from the sociocultural theory of mind as elaborated by Russian psychologist L. S. Vygotsky. By offering mediation when individuals experience difficulties and carefully tracing their responsiveness, Vygotsky (1998) proposed that diagnoses may uncover abilities that have fully formed as well as those still in the process of developing. This insight has led to numerous assessments, collectively referred to as DA, that have been pursued primarily in the domains of special education and general cognitive abilities measurement (Feuerstein, Feuerstein, & Falik, 2010; Haywood & Lidz, 2007). To date, L2 DA work has been primarily conducted in classroom settings (Ableeva, 2010; Lantolf & Poehner, 2011; Poehner, 2007, 2008). This paper discusses a recent project concerning the design of online multiple-choice tests of L2 reading and listening comprehension that leverage the principle that mediation is indispensable for diagnosing development. Specifically, each test item is accompanied by a set of prompts graduated from implicit to explicit. In this way, resultant diagnoses include not only whether learners answered correctly (their actual score) but also the amount of support they required (mediated score) during the test. We argue that the set of scores automatically generated by the tests, together with a breakdown of learner performance on items targeting particular component features of comprehension, provide a fine-grained diagnosis of their L2 development while also offering information relevant to subsequent teaching and learning.