The TEAM layered architecture provides a solution for realizing a high level of mission autonomy for a group of heterogenous autonomous underwater vehicles (AUVs) that collectively work towards a common goal. The architecture is particularly suited for managing systems that need to operate in communications-limited environments such as the underwater environment [1]. TEAM minimizes the requirements for communication by using a layered architecture following 4D/RCS principles [2], where missions are hierarchically decomposed into tasks and each layer has its own control functionality which is responsible for the planning, execution and evaluation of tasks. The top layer thus delegates the planning, execution and evaluation of tasks to a lower layer, requesting that the specified performance goals are met. This paper details the design choices and implementation of TEAM, explaining the ideas on which the architecture is based and providing an anchor for publishing subsequent work. The paper will also show some results obtained during real life demonstrations performed during the REPMUS 2021 experiments [3]. The general concepts implemented in TEAM can be applied to many autonomous applications requiring collaborative mission autonomy. The layered architecture provides an interesting view on how to approach a complex autonomy problem by breaking the problem into smaller pieces, and using vehicle autonomy within a cooperative framework.