Mobile ad-hoc networks (MANETs) are the best choice for instant networks when no fixed infrastructure is available. MANETs can support applications in a variety of areas like emergency assistance and inter-vehicle communications. Most developed wireless ad-hoc routing protocols are designed to discover and maintain an active path from source to destination with an assumption that every node is friendly. However, it is possible that the participating nodes may be selfish or malicious. A mechanism to evaluate reputation and trust for each node is essential for the reliability of routing protocol in MANETs. We integrate reputation and trust management into routing protocols in MANETs. Reputation mechanism is based on constantly monitoring and updating first-hand and second-hand information. The nodes within the network are able to monitor their neighbors and obtain first-hand information based on the perceived behavior. Second-hand information is obtained from the sharing of first-hand information with other nodes. The nodes thus create total reputation value by a combination of firsthand and second-hand information. The total reputation value is then available to neighboring nodes for routing decisions. Dirichlet distribution is combined with Bayes theorem to provide finer granularity for nodes classification based on their behavior. Fine-grained reputation management is integrated into a routing protocol to explore the possibility and benefits on improvement of reliability in both route discovery and maintenance in MANETs.