Hi,
in general you have both options.
1) You can formulate material model directly in terms of forces and moments and strains and curvatures. Typically this is done for homogeneous cross sections (made of single material) but I have also seen these kind of models for concrete + reinforcement, then you need additional parameters determining position and properties of reinforcing bars, for example. This should be implemented as new material model, thus you can store history variables in associated status.
2) You can opt for slightly more refined model of cross section by regarding it as a section made of parallel layers (or fibres in 3d), where one can assign dimension and position to each layer as well as unique material model to each layer. The strains in layers are determined by beam kinematics. Then the response (beams expect to get forces and moments) is obtained by integration on the cross section level. The cross section introduces special integration rule to make this happen by looping over individual layers. Requires to develop new material model supporting stress strain state in layer. There are integration points for each layer, and material model for each layer can also store its history there.
There is working implementation of this concept (layered and fibred cross section models).  On the other hand, to account for inclined bars or stirrups would is somehow difficult. In this case I would go to 2d/3D model.

This should work to access sectional properties from a material

StructuralCrossSection *crossSection = static_cast< StructuralCrossSection * > ( gp->giveElement()->giveCrossSection());                                        
double area = crossSection->give(CS_Area, gp);
double Iy   = crossSection->give(CS_InertiaMomentY, gp);