The existing architecture for intent-based systems consists of four blocks: Intent translation, resolution, activation, and assurance. Such architecture, even though straightforward for small problems, results in prohibitively complex to maintain codebase for larger problems. This is due to mixing three distinct aspects of autonomous systems: optimal control, decision support functionality, and game-theory aspects. To handle this complexity, partitioning and agent-based approaches are used. These help to some extent but either directly result in sub-optimality or rapidly lead to complexity-induced stagnation of the intent system. Moreover, optimization and control theory aspects are afterthoughts. We propose an architecture that is based on optimal control abstracting the system (i.e., plant) for a decision support system which in turn consults a game theory core. This way, almost all decision theory questions are isolated in the optimal control core and game theory questions in game theory core. This separation of concerns enables agnostic algorithms capable of handling many users, intents, and components, along with their dynamic changes, optimally and without requiring changes in the codebase of the intent system.