Curbing greenhouse gas emissions is a challenge of the utmost importance for our society future and requires urgent decisions on the implementation of clear-cut climate economic policies. Integrated Assessment Models (IAMs) allow to explore alternative energy scenarios in the next 30-70 years. They are key to support the design of climate policies as they highlight the nexus between climate modelling, social science, and energy systems. However, the use of IAMs to inform climate policies does not come free of controversial aspects. Primarily, the inherent uncertainty of IAMs long-term outputs has created several difficulties for the integration of the modelling insights in the policy design. Modelling outputs diverge across IAMs models quite dramatically when they are asked for example to quantify the uptake of key technologies for the decarbonisation, such as renewables and carbon capture and storage. Uncertainty in IAMs descends from lack of knowledge of the future and from IAMs incomplete representations of the future. Uncertainty cannot be removed, but reduced, understood, and conveyed appropriately to policy makers to avoid that different projections cause delayed actions. This project aims to fill this gap providing a methodology which defines the sources of uncertainty, either due to IAMs inputs or IAMs structure, and quantify their relative importance. The methodology will be embodied in an emulator of IAMs, MANET (the eMulAtor of iNtegratAd assEssmenT models) formulated using machine learning techniques to reproduce IAMs outputs. The project will provide a proof of concept of MANET focusing on the uptake of key decarbonisation technologies. The emulator will provide a simplified version of the IAM outputs as a response surface of the model to any variation of the inputs. MANET will be a flexible tool for policy makers and scientists for a direct comparison of IAMs with no limitation of the solution domain.