How to estimate error propagation on regressed function from covariance matrix using scipy.curve_fit?

I need a hand with something. I'm using curve_fit from SciPy to fit a curve with three parameters. Now, when I use it, I get this thing called pcov, which I know is the covariance matrix. But here's the thing: can I use pcov to figure out how much my predictions might be off?

I know I can calculate the error between the real data and the fitted curve, but can I somehow use pcov to get that error variance directly for a new data point I'm predicting?

Thanks in advance