Studying proteomes within subcellular structures presents significant challenges, particularly for structures that are membrane-less or cannot be isolated from cells. Addressing this hurdle, MicroscoopTM emerges as a groundbreaking technology capable of precisely capturing proteins at specified subcellular regions of interest (ROIs) with nanoscale precision. This sophisticated system integrates microscopy, optics, mechatronics, photochemistry, and deep learning to enable high-content in situ photolabeling. With MicroscoopTM, proteins are precisely biotinylated within user-defined cellular organelles, granules, or cell-cell contact surfaces under a microscope, utilizing directed photochemistry within each field of view (FOV). This process is automatically repeated across thousands of FOVs to photo-label cellular structures sharing similar morphological features. Subsequently, ample biotinylated proteins are obtained for streptavidin pulldown and mass spectrometry analysis. The robustness of this approach is demonstrated through the successful mapping of the human cellular nucleus proteome, identifying over 1000 nuclear proteins with a specificity exceeding 90%. Further analysis reveals comprehensive coverage of nuclear complexes and identification of low-copy-number proteins. In one study using MicroscoopTM technology to explore the nucleolus proteome, 97 of the top 100 abundant proteins are confirmed to be originated from the nucleolus. In another investigation focusing on stress granules (SG) proteome, we identified 2,614 proteins, including 200 with low copy numbers. Although the specificity is moderate, with only 66% of the top 50 abundant proteins mapped with the known stress granules database, functional enrichment analysis highlights 13 non-SG associated proteins as high-confidence core interactors within the stress granule network. Among them, 11 proteins (PDLIM7, EIF3CL, YWHAE, RPSA, MTA2, UGDH, DDX17, ANLN, PSMD3, PSMA6, and MCM2) are confirmed to co-localize with G3BP1 through immunostaining, thus elevating the specificity eventually to 96% SG association. Overall, our study illustrates that MicroscoopTM enables hypothesis-free, comprehensive mapping of subcellular proteomes at user-defined regions of interest, significantly advancing cell biology by revealing new proteins or biomarkers.