Recognizing the critical role of teams in the U.S. military, the Department of Defense has been a strong supporter of research into team training and performance. In this paper we will describe a newly developed simulation-based testbed suitable for supporting a wide variety of programs within this research area. The freely-available software suite includes: (a) a scenario authoring tool, (b) a simulation testbed for a complex, three-role team task, (c) three synthetic agents, each capable of playing one of the roles in the testbed, and (d) tools for performance recording and assessment. The functional specifications of the software suite were developed collaboratively by a group of industrial engineers, computer scientists, and research psychologists. The task is a complex, dynamic team task designed to simulate air defense operations. The three roles are distinct, with separate interfaces and responsibilities, and are organized into a simple command hierarchy of one leader and two supporting team members. Scenarios created in the authoring tool are deterministic, allowing for a high degree of experimental control. Scenario difficulty can be controlled along multiple dimensions, including raw data volume, resource allocation complexity and information ambiguity. The synthetic agents may be used as teammates for human participants during data collection. These agents are capable of performing competently on previously unencountered scenarios within a broad (but not infinite) problem space. In addition, the experimenter may degrade the agents' performance on an event-by-event basis through flags set in the scenario authoring tool. The testbed captures and records multi-modal performance data at multiple levels of detail ranging from time-stamped keystrokes, mouse actions and speech interactions to an analysis of actions within a "Windows of Opportunity" assessment framework. Cooperative use of this testbed within government, academia, and industry will generate a more reliable and valid understanding of team training and performance.