Abstract Background Breast cancer is the second leading cause of cancer deaths among females worldwide. Surgical resection is the most common treatment option for women diagnosed with breast cancers. Many women with in situ disease or early invasive breast cancer are candidates for breast conserving surgery (BCS), which involves removing the lesion of interest and preserving the rest of the breast. One of the greatest challenges a breast cancer surgeon faces is to achieve negative margins while optimizing aesthetic outcomes. Negative margins are highly desirable as they offer the greatest potential for prolonged disease-free survival. Accurate intraoperative surgical margin evaluation is therefore critical during BSC. Molecular technologies offer the opportunity to incorporate cancer-specific biomarkers into clinical decision-making for improved cancer detection. Our team has reported the development of an innovative handheld mass spectrometry (MS) probe, the MasSpec Pen (MSPen), for rapid and gentle molecular analysis of human tissues, including breast cancer tissues. The MSPen uses water to gently extract molecules from tissues upon contact, which are then analyzed by MS and statistical classifiers to provide a predictive classification within seconds. Here, we evaluate the performance of the MSPen for intraoperative detection of breast cancer. Methods The MSPen coupled to an Orbitrap Exploris 120 (Thermo Scientific) is being used to analyze both normal breast and invasive lobular carcinoma (ILC) banked tissues. Tissue samples are first thawed prior to analysis. After analysis, the region analyzed is demarcated, and the samples snap-frozen, sectioned, H&E stained, following evaluation by a pathologist. We will then apply the least absolute shrinkage and selector operator (lasso) method to build classification models for distinguishing normal breast and ILC using the histologically validated mass spectra. The MSPen system is also implemented in the operating room for testing within surgical workflow. Analyses will be carried out during BCS surgeries for patients with ILC. The ILC classifier generated from banked tissues will be used to predict on data obtained from the intraoperative analyses and validated against final histopathological reports. Results To date, we have analyzed 213 banked human breast tissues using the MSPen, including healthy breast (n = 79), invasive ductal carcinoma (IDC) (n = 64), healthy lymph nodes (n = 26), and lymph nodes with metastatic IDC (n = 44). The rich mass spectral data obtained from breast tissues showed high relative abundance of various metabolites and lipid species that have been previously described as potentially diagnostic of breast cancer. Using this data, we built classifiers that discriminate cancer and normal breast tissues with high accuracy (overall 94%) for primary breast IDC. In a pilot clinical study, we demonstrated feasibility for ex vivo and in vivo tissue analysis using the MSPen platform during 29 breast surgeries. We are currently expanding our study to detect ILC using banked tissues and in patient tissues intraoperatively. Conclusion Our study in breast cancer provides evidence that the MSPen enables molecular-based diagnosis of IDC tissues using excised specimens and in in vivo and ex vivo analysis performed by surgeons intraoperatively. Current effort including ILC is expected to show similar capabilities for patients with ILC.