The sign language phenomenon that some scholars refer to as “agreement” has triggered controversial discussions among sign language linguists. Crucially, it has been argued to display properties that are at odds with the notion of agreement in spoken languages. A thorough theoretical investigation of the phenomenon may thus add to our understanding of the nature and limits of agreement in natural language. Previous analyses of the phenomenon can be divided into three groups: (i) gesture-based non-syntactic analyses, (ii) hybrid solutions combining syntactic and semantic agreement, and (iii) syntactic accounts under which agreement markers are reanalyzed as clitics. As opposed to these accounts, we argue in this paper that sign language agreement does represent an instance of agreement proper, as familiar from spoken language, that is fully governed by syntactic principles. We propose an explicit formal analysis couched within the Minimalist Program that is modality-independent and only involves mechanisms that have been independently proposed for the analysis of agreement in spoken language. Our proposal is able to capture the (apparent) peculiarities of sign language agreement such as the distinction of verb types (only some verbs show agreement), the behavior of backwards verbs (verbs displaying agreement reversal), and the distribution of the agreement auxiliary. However, we suggest that the combination of mechanisms is modality-specific, that is, agreement in sign language, and in German Sign Language in particular, involves modality-independent ingredients, but uses a modality-specific recipe which calls for a (somewhat) unusual combination of independently motivated mechanisms.