This paper develops a strictly Agree-based analysis of the direct/inverse agreement pattern in Guarani. Adopting a model of Agree based on Cyclic Agree (Bejar 2003, Bejar and Rezac 2009), it further follows (Bejar 2003) and (Bejar and Rezac 2009) in that the crucial syntactic distinction between direct and inverse agreement is how many arguments the probe establishes Agree with. Extending this to Guarani, the proposal is that the direct is the result of single Agree between one probe and one goal, and the inverse is the result of double Agree between one probe and two separate goals. Double Agree gives rise to a unique probe structure with two values (one from each DP). In the morphology, whichever Vocabulary Item (VI) is to be inserted must express both. It is only the Guarani inverse VIs which are capable of expressing two values on a single probe. This also helps explain the presence of direct agreement in intransitives. A secondary # probe is shown to give rise to the portmanteau in local direct scenarios.