MASON is a Java-based multi-agent simulation toolkit designed for a wide range of applications, including swarm robotics, machine learning, and social complexity environments. It provides a flexible and extensible framework for simulating multi-agent systems, with a clear separation between the model and visualization components. This allows models to be detached from or attached to visualizers, and to run on different platforms during execution. The system is open-source and free, developed by George Mason University's Computer Science Department and the Center for Social Complexity. MASON is designed to be fast, portable, and capable of producing guaranteed duplicable results. It supports both 2D and 3D visualization, and allows for efficient simulation of up to a million agents without visualization. The system is also capable of checkpointing simulations to disk, enabling them to be resumed on any platform. MASON's architecture is built in three layers: utility, model, and visualization. The model layer contains the core simulation logic, including a discrete-event schedule, random number generator, and fields for storing objects and locations. The visualization layer provides GUI-based visualization and manipulation of the model. MASON has been used in various applications, including cooperative target observation, ant foraging, urban traffic simulation, and anthrax propagation. These applications demonstrate the toolkit's broad applicability and versatility. MASON is also capable of being embedded into larger existing libraries and supports a wide range of features, including network modeling, charting, and parameterization. Compared to other simulation environments, MASON offers a unique combination of architecture and features, particularly in its ability to separate the model from visualization dynamically. It is also faster than many other multi-agent simulation libraries and is designed to produce duplicable results. MASON has been used to port existing simulations from other environments, such as Ascape and SWARM, to take advantage of its features and capabilities. The system is available as open-source under a BSD-style license and can be accessed at http://cs.gmu.edu/~eclab/projects/mason/.MASON is a Java-based multi-agent simulation toolkit designed for a wide range of applications, including swarm robotics, machine learning, and social complexity environments. It provides a flexible and extensible framework for simulating multi-agent systems, with a clear separation between the model and visualization components. This allows models to be detached from or attached to visualizers, and to run on different platforms during execution. The system is open-source and free, developed by George Mason University's Computer Science Department and the Center for Social Complexity. MASON is designed to be fast, portable, and capable of producing guaranteed duplicable results. It supports both 2D and 3D visualization, and allows for efficient simulation of up to a million agents without visualization. The system is also capable of checkpointing simulations to disk, enabling them to be resumed on any platform. MASON's architecture is built in three layers: utility, model, and visualization. The model layer contains the core simulation logic, including a discrete-event schedule, random number generator, and fields for storing objects and locations. The visualization layer provides GUI-based visualization and manipulation of the model. MASON has been used in various applications, including cooperative target observation, ant foraging, urban traffic simulation, and anthrax propagation. These applications demonstrate the toolkit's broad applicability and versatility. MASON is also capable of being embedded into larger existing libraries and supports a wide range of features, including network modeling, charting, and parameterization. Compared to other simulation environments, MASON offers a unique combination of architecture and features, particularly in its ability to separate the model from visualization dynamically. It is also faster than many other multi-agent simulation libraries and is designed to produce duplicable results. MASON has been used to port existing simulations from other environments, such as Ascape and SWARM, to take advantage of its features and capabilities. The system is available as open-source under a BSD-style license and can be accessed at http://cs.gmu.edu/~eclab/projects/mason/.