The challenge of integrated exhaust consistent with the other requirements in DEMO and power plant class tokamaks (ITER-like and alternative DEMOs, FNSF approaches) is wellknown and the exhaust solution is likely to be fundamental to the design and operating scenarios chosen. Strategies have been proposed such as high main plasma radiation (e.g. [1]), but improved solutions are sought and will require revised research methodologies. While no facility can address all the challenges, the new MAST Upgrade tokamak enables exploration of a wide range of divertor plasma aspects in a single device and their relation with the core plasma (e.g. access to H-mode), in particular the development of fundamental understanding and new ideas. It has a unique combination of closed divertor, capability of a wide range of configurations from conventional to long-leg (including Super-X), and fully symmetric double null (plasma and divertor structures). To extrapolate to DEMO and power plant scale devices where full integrated tests in advance are not feasible yet different physics mechanisms may dominate, theory-based models are likely to be essential, for confident performance prediction, optimisation, and a "qualification" of the concept. Development and validation of such models is at the heart of the programme around MAST Upgrade. Amongst the many areas to be explored, there will be a strong focus on the closely coupled topics of plasma detachment and cross-field transport mechanisms (e.g. plasma filaments), key ingredients of effective and reliable protection of the plasma facing components at DEMO-scale. Index Terms-Fusion reactor design, divertor, plasma exhaust, plasma filaments, super-X, tokamak devices. I. INTRODUCTION-ALTERNATIVE EXHAUST APPROACHES Tokamaks at reactor scale need effective and practical exhaust systems. An integrated exhaust solution accommodates both a high performance plasma and the engineering and materials requirements of reliably protected long-lifetime plasma-facing components (PFCs). Its many challenges are well documented-it involves far more than the divertor configuration. The fastest path would be to use the single-null divertor configuration (e.g. as implemented on ITER) accompanied by a highly radiating main plasma and a fully detached divertor [1]. However, while there is some basis for optimism, it is not yet clear whether such a constraint on the