Compatibility, interoperation and tooling are important problems facing new language implementations. On the Truffle/Graal platform, there is a recurring need for solutions to these problems particularly in /VM-to-native/ and /cross-VM/ scenarios. I’ll present my recent work on liballocs, and outline how it offers one possible approach for conveniently tackling these problems. The core service of liballocs is a meta-level API which allows viewing an entire Unix process at the level of typed objects, explicitly modelling the open-ended collection of diverse allocators, compilers and loaders that it might contain. I’ll describe briefly how this already works for a large fraction of real C code, and extends easily to C++, Fortran et al. Then we’ll discuss requirements and approaches for extending this to Truffle languages – including basic native code interoperation, but also how it might work between distinct high-level languages or even across distinct VMs.