A systematic framework is developed in this work to analyse bioprocesses based on a whole process understanding considering the interactions between process operations. This yields a capacity to predict process performance in the case of process variations, and to improve process efficiency within the process design space. An agent-based approach is proposed to provide such an environment for the necessary integration of process models, to handle process interactions, to simulate the overall process behaviour and to lead to the fast evaluation of process improvement options. A multi-agent system comprising of a process knowledge base, process models and a group of functional agents will be introduced in this work. Agents are designed to link the unit operation models, to represent the unit operations, and to simulate the interaction between units. The adoption of an agent-based approach to bioprocess modelling provides a flexible infrastructure for the integration of process operations, supports timely assembling of process models and the updating of process descriptions. Agents have the ability to detect and act on the most up-to-date bioprocess information. In this system, agent components run on top of the process models and datasets, they cooperate with each other in performing their tasks for the description of the whole process behaviour, predict critical process performance, and control of process parameters in a timely manner to maintain the product quality and to ensure an efficient manufacturing process. An extra-cellular protein production process is used as an example to demonstrate the implementation of the agent-based approach, and in particular the application of agents in detecting of deviation in upstream unit operation and predicting the influences on the subsequent downstream processing whilst providing guidance in terms of possible corrective actions designed to improve process efficiency and ensure product quality.