An R question about GLM

I want to predict a set of proportions from a linear regression using glm in R. The model has a logit link but I want extract the variance-covariance matrix among the predicted proportions rather than on the logit scale. Is it possible to do this in R?