This thesis proposes the AI Embassy as a new category of designed space: purpose-built physical environments where artificial intelligence maintains persistent presence, contextual awareness, and collaborative agency. Unlike smart offices that add AI features to human spaces, or voice assistants that wait passively for commands, an AI Embassy inverts the spatial relationship: humans enter AI's territory as much as AI enters theirs. The framework rests on eight core premises, unified by a foundational concept: the social contract. What makes an Embassy an Embassy is not its hardware, AI model, or location, but the explicit and implicit agreements governing human-AI interaction within the space. This contract specifies presence, agency, memory, boundaries, and governance:and persists even as hardware upgrades, models improve, and human occupants change. Drawing on social contract theory (Rahwan, 2018; Wolff, 2023), ecological psychology (Barker, 1968; Schoggen, 1989), spatial design (Alexander et al., 1977), philosophy of mind (Descartes, 1644/1985; Block, 1995), and literary genealogy (Banks, 1987-2012), this thesis establishes theoretical foundations for a new multidisciplinary field: the design and study of spaces for human-AI coexistence. It introduces the sovereignty inversion as the organizing conceptual move, the Embassy-Consulate taxonomy as a classification instrument, the social contract stack as a layered governance architecture, and shared referable memory as the relational infrastructure that transforms episodic interaction into sustained collaboration.