ErbB2 expression in early breast cancer can predict tumor aggressiveness and clinical outcomes in large patient populations. Accurate assessment with physical biopsy and conventional pathology can be limited by tumor heterogeneity. We aim to demonstrate real-time optical sectioning using a near-infrared labeled ErbB2 peptide that generates tumor-specific contrast in human xenograft breast tumors in vivo. We used IRDye800CW as the fluorophore, validated performance characteristics for specific peptide binding to cells in vitro, and investigated peak peptide uptake in tumors using photoacoustic tomography. We performed real-time optical imaging using a handheld dual-axes confocal fluorescence endomicroscope that collects light off-axis to reduce tissue scattering for greater imaging depths. Optical sections in either the vertical or horizontal plane were collected with sub-cellular resolution. Also, we found significantly greater peptide binding to pre-clinical xenograft breast cancer in vivo and to human specimens of invasive ductal carcinoma that express ErbB2 ex vivo. We used a scrambled peptide for control. Peptide biodistribution showed high tumor uptake by comparison with other organs to support safety. This novel integrated imaging strategy is promising for visualizing ErbB2 expression in breast tumors and serve as an adjunct during surgery to improve diagnostic accuracy, identify tumor margins, and stage early cancers.