Affinity reagents recognizing biomarkers specifically are essential components of clinical diagnostics and target therapeutics. However, conventional methods for screening of these reagents often have drawbacks such as large reagent consumption, the labor-intensive or time-consuming procedures, and the involvement of bulky or expensive equipment. Alternatively, microfluidic platforms could potentially automate the screening process within a shorter period of time and reduce reagent and sample consumption dramatically. It has been demonstrated recently that a subpopulation of tumor cells known as cancer stem cells possess high drug resistance and proliferation potential and are regarded as the main cause of metastasis. Therefore, a peptide that recognizes cancer stem cells and differentiates them from other cancer cells will be extremely useful in early diagnosis and target therapy. This study utilized M13 phage display technology to identify peptides that bind, respectively, to colon cancer cells and colon cancer stem cells using an integrated microfluidic system. In addition to positive selection, a negative selection process was integrated on the chip to achieve the selection of peptides of high affinity and specificity. We successfully screened three peptides specific to colon cancer cells and colon cancer stem cells, namely, HOLC-1, HOLC-2, and COLC-1, respectively, and their specificity was measured by the capture rate between target, control, and other cell lines. The capture rates are 43.40 ± 7.23%, 45.16 ± 7.12%, and 49.79 ± 5.34% for colon cancer cells and colon cancer stem cells, respectively, showing a higher specificity on target cells than on control and other cell lines. The developed technique may be promising for early diagnosis of cancer cells and target therapeutics.