Multiphysics simulations are paramount in most applications and often occur through the coupling of various physics in different physical domains coupled through surfaces in 3D. Relying on code heritage, it is desirable to design a coupling strategy that allows to reuse existing model-specific solvers with minimal modifications. However, in most existing numerical strategies, the design of such strategy comes with low-order coupling and difficulties in order to control the errors related to the coupling strategy. In this contribution, we introduce a novel strategy in order to ensure adaptive accurate and stable coupled simulations. The principle is a multistep coupling scheme, which relies on the history of the exchanged quantities to enable a high-order accurate coupling with time adaptation and error control. Explicit and implicit variants are discussed. Numerical experiments conducted with an open-source demonstrator on a conjugate heat transfer problem show that high-order convergence can be reached, and that stability is favorable compared to other classical approaches. We propose some insight on the numerical analysis of the proposed strategy and extensions to simulations in combustion are discussed.