In this dissertation, I provide a game-theoretic analysis of three models of strategic communication and mechanism design. In Chapter 1, I analyze cheap talk communication under the receiver's uncertainty about the sender's expertise. I find that in a one-shot cheap-talk communication game, the informativeness of the sender's messages need not increase in the receiver's prior belief about the sender's expertise. Using this result, I further show that in a two-period repeated game, the sender's reputation concerns might force her to report less truthfully, thus making communication less efficient from the receiver's point of view. In Chapter 2, I consider the matching platform's problem of implementing an incentive compatible and individually rational matching mechanism. The players on two sides have their private types that determine their outside options and their value to their potential matches. The platform's problem is to (i) specify a matching function that depends on players' reports and (ii) a transfer function. For a general finite-player setting, I characterize implementable mechanisms. Using this characterization, I show that a random matching mechanism is always implementable. Using a constructive proof, I also show that implementability of positive assortative matching is possible but not guaranteed. In Chapter 3, I study a model of information sharing in which two firms have to take actions in an uncertain environment. They both have private signals and would benefit from learning each other's signals. Yet sharing the signal hurts each of them while increasing the overall efficiency. I find that introducing an intermediary who offers information-sharing contracts improves efficiency of communication. If firms have signals of similar precision then full sharing is possible using these contracts. If precision is unequal, then the less informed firm shares fully and the more informed firm shares its information partially.