Background Immunotherapy is revolutionizing cancer treatment from conventional radiotherapies and chemotherapies to immune checkpoint inhibitors which use patients’ immune system to recognize and attack cancer cells. Despite the huge clinical success and vigorous development of immunotherapies, there is a significant unmet need for a robust tool to identify responders to specific immunotherapy. Early and accurate monitoring of immunotherapy response is indispensable for personalized treatment and effective drug development.Methods We established a label-free metabolic intravital imaging (LMII) technique to detect two-photon excited autofluorescence signals from two coenzymes, NAD(P)H (reduced nicotinamide adenine dinucleotide (phosphate) hydrogen) and FAD (flavin adenine dinucleotide) as robust imaging markers to monitor metabolic responses to immunotherapy. Murine models of triple-negative breast cancer (TNBC) were established and tested with different therapeutic regimens including anti-cluster of differentiation 47 (CD47) immunotherapy to monitor time-course treatment responses using the developed metabolic imaging technique.Results We first imaged the mechanisms of the CD47-signal regulatory protein alpha pathway in vivo, which unravels macrophage-mediated antibody-dependent cellular phagocytosis and illustrates the metabolism of TNBC cells and macrophages. We further visualized the autofluorescence of NAD(P)H and FAD and found a significant increase during tumor growth. Following anti-CD47 immunotherapy, the imaging signal was dramatically decreased demonstrating the sensitive monitoring capability of NAD(P)H and FAD imaging for therapeutic response. NAD(P)H and FAD intravital imaging also showed a marked decrease after chemotherapy and radiotherapy. A comparative study with conventional whole-body bioluminescence and fluorescent glucose imaging demonstrated superior sensitivity of metabolic imaging. Flow cytometry validated metabolic imaging results. In vivo immunofluorescent staining revealed the targeting ability of NAD(P)H imaging mainly for tumor cells and a small portion of immune-active cells and that of FAD imaging mainly for immunosuppressive cells such as M2-like tumor-associated macrophages.Conclusions Collectively, this study showcases the potential of the LMII technique as a powerful tool to visualize dynamic changes of heterogeneous cell metabolism of cancer cells and immune infiltrates in response to immunotherapy thus providing sensitive and complete monitoring. Leveraged on ability to differentiate cancer cells and immunosuppressive macrophages, the presented imaging approach provides particularly useful imaging biomarkers for emerged innate immune checkpoint inhibitors such as anti-CD47 therapy.