Local recurrence is a common cause of treatment failure for patients with solid tumors. Intraoperative detection of microscopic residual cancer in the tumor bed could be used to decrease the risk of a positive surgical margin, reduce rates of reexcision, and tailor adjuvant therapy. We used a protease-activated fluorescent imaging probe, LUM015, to detect cancer in vivo in a mouse model of soft tissue sarcoma (STS) and ex vivo in a first-in-human phase 1 clinical trial. In mice, intravenous injection of LUM015 labeled tumor cells, and residual fluorescence within the tumor bed predicted local recurrence. In 15 patients with STS or breast cancer, intravenous injection of LUM015 before surgery was well tolerated. Imaging of resected human tissues showed that fluorescence from tumor was significantly higher than fluorescence from normal tissues. LUM015 biodistribution, pharmacokinetic profiles, and metabolism were similar in mouse and human subjects. Tissue concentrations of LUM015 and its metabolites, including fluorescently labeled lysine, demonstrated that LUM015 is selectively distributed to tumors where it is activated by proteases. Experiments in mice with a constitutively active PEGylated fluorescent imaging probe support a model where tumor-selective probe distribution is a determinant of increased fluorescence in cancer. These co-clinical studies suggest that the tumor specificity of protease-activated imaging probes, such as LUM015, is dependent on both biodistribution and enzyme activity. Our first-in-human data support future clinical trials of LUM015 and other protease-sensitive probes.