Thrust and cudaLaunchCooperativeKernel

How can I use a thrust::device_vector as a parameter to cudaLaunchCooperativeKernel?

like usage with most other kernel launches I can think of, you would not use a thrust::device_vector object directly, but would instead extract a pointer or iterator of some sort to the underlying data, and use that, much as you would typically use pointers to data for CUDA kernel launches.