Mimetics of conformational protein epitopes have broad applications but have been difficult to identify using conventional peptide phage display. The 10th type III domain of human fibronectin (FNfn10) has two extended, randomizable surface-exposed loops and might be more amenable to the identification of such mimetics. We therefore selected a library of FNfn10 clones, randomized in both loops (15 residues in all), for binding to monoclonal antibodies (mAbs) that recognize the HIV-1 envelope glycoprotein. Anti-idiotypic monobodies (αIMs) mimicking both "linear" epitopes (2F5 and 4E10 mAbs) and conformational epitopes (b12 and VRC01 mAbs) were generated. αIMs selected against 2F5 and 4E10 frequently displayed sequence homology to the corresponding linear native epitopes. In the case of b12 and VRC01, we expected that the two constrained loop domains of FNfn10 would both contribute to complex conformational interactions with target antibodies. However, mutagenesis studies revealed differences from this simple model. An αIM selected against b12 was found to bind its cognate antibody via only a few residues within the BC loop of FNfn10, with minimal contribution from the FG loop. Unexpectedly, this was sufficient to generate a protein that engaged its cognate antibody in a manner very similar to that of HIV-1 Env, and with a strong KD (43 nM). In contrast, an αIM selected against VRC01 engaged its cognate antibody in a manner that was dependent on both BC and FG loop sequences. Overall, these data suggest that the FNfn10 scaffold can be used to identify complex structures that mimic conformational protein epitopes.