This paper explores empirical merits of a version of Agree that is defined based on Minimal Search (MS-Agree). Compared to the standard Agree, MS-Agree, essentially a search algorithm, uniquely allows the independent assignment of its search target and search domain. This unique feature enables MS-Agree to accommodate both upward and downward agreement phenomena, and offers a unified downward search analysis for negative concord, inflection doubling, multiple case-assignment, cyclic agreement, and complementizer agreement observed across languages. This paper thus argues that these core empirical data that have served as the main motivation for Upward Agree can be successfully reanalyzed with MS-Agree. It is also argued that the proposed MS-Agree analysis makes better predictions than Upward Agree regarding intervention effects in apparent upward agreement phenomena.