Abstract In a mechanism, a designer may reveal some information to influence agents’ private types in order to obtain more payoffs. In the literature, the information is usually represented as random variables, the value of which are realized by the nature. However, this representation of information may not be proper in some practical cases. In this paper, we propose a type-adjustable mechanism where the information sent by the designer is modeled as a solution of her optimization problem. From the designer’s perspective, the probability distributions of agents’ private types may be optimally controlled. By constructing a type-adjustable first-price sealedbid auction, we show that the seller may obtain more expected payoffs than what she could obtain at most in the traditional optimal auction model. Interestingly, to the satisfaction of all, each agent’s ex-ante expected payoffs may be increased too. In the end, we compare the type-adjustable mechanism with other relevant models. Key words: Mechanism design; Optimal auction; Bayesian implementation.