Multiply sectioned Bayesian networks (MSBNs) are an extension of Bayesian networks for flexible modeling and cooperative multiagent probabilistic inference. A large and complex equipment is modeled by a set of Bayesian subnets in a MSBN each of which corresponds to a natural component of the equipment. Each subnet forms the core knowledge of an autonomous M&D agent. Inference is performed in a distributed fashion while answers to queries are coherent with respect to probability theory. Recent advance in the MSBN theory shows that the coherence is not compromised even when internal knowledge of each agent is kept private from on another. Hence M&D systems can be integrated for very large equipments from agents built by different vendors. We overview the MSBN framework and its features relevant to M&D tasks. We illustrate applications of MSBNs to M&D using constructed examples. We discuss several ongoing research issues regarding internet based MSBNs for M&D, agent upgrading through learning and handling of dynamic systems.