Since Bellare and Rogaway's work in 1994, the indistinguishability-based security models of authenticated key agreement protocols in simple cases have been evolving for more than ten years. In this paper, we review and organize the models under a unified framework with some new extensions. By providing a new ability (the Coin query) to adversaries and redefining two key security notions, the framework fully exploits an adversary's capacity and can be used to prove all the commonly required security attributes of key agreement protocols with key confirmation. At the same time, the Coin query is also used to define a model which can be used to heuristically evaluate the security of a large category of authenticated protocols without key confirmation. We use the models to analyze a few identity-based authenticated key agreement protocols with pairings.