Current electronic commerce applications such as online auction systems are not trustworthy due to a lack of effective trust management mechanisms. A trustworthy online auction system requires a dynamic trust management module that can detect abnormal bidding activities in real-time, notify the involved users, and cancel the corresponding auction immediately. In this paper, we present a general framework for agent-based trust management (ATM) in online auctions. The ATM module consists of three types of agents, namely the monitoring agent, the analysis agent and the security agent. A monitoring agent can monitor a bidder and detect any abnormal bidding behavior; while the analysis agent and the security agent can analyze state-based information and history information of a bidder, and make decisions on shill detection, respectively. We illustrate the communication protocol among various agents, and demonstrate our agent-based trust management approach for online auctions using a prototype ATM module developed with JADE.