Abstract Establishing information exchange between existing knowledge-based systems can lead to devastating inconsistency. Automatic resolution of inconsistency often is unsatisfactory, because any modification of the information flow may lead to bad or even dangerous conclusions. Methods to identify and select preferred repairs of inconsistency are thus needed. In this work, we leverage the expressive power and generality of Multi-Context Systems (MCS), a formalism for information exchange, to select most preferred repairs, by use of a meta-reasoning transformation. As for computational complexity, finding preferred repairs is not higher than the base case; finding most-preferred repairs is higher, yet worst-case optimal.