You can def add your own mean function or whatever you’d like there. Something like
mean = pm.gp.mean.Linear(...)
mogp = pm.gp.Marginal(mean_func=mean, cov_func=cov_icm)
(taking code from @daniel-saunders-phil’s link) might be what you need. It’s important to know that the mean function is just a super simple wrapper around any PyMC code, so you can write your own easily. __init__ takes in random variables, and __call__ takes in the same data X the GP is over.