The disruption of professional childcare has emerged as a substantial collateral consequence of the public health precautions related to COVID-19. Increasingly, it is becoming clear that childcare centers must be (at least partially) operational in order to further mitigate the socially debilitating challenges related to pandemic induced closures. However, proposals to safely reopen childcare while limiting COVID-19 outbreaks remain understudied, and there is a pressing need for evidence-based scrutiny of the plans that are being proposed. Thus, in order to support safe childcare reopening procedures, the present study employed an agent-based modeling approach to generate predictions surrounding risk of COVID-19 infection and student-days lost within a hypothetical childcare center hosting 50 children and educators. Based on existing proposals for childcare and school reopening in Ontario, Canada, six distinct room configurations were evaluated that varied in terms of child-to-educator ratio (15:2, 8:2, 7:3), and family clustering (siblings together vs. random assignment). The results for the 15:2 random assignment configuration are relevant to early childhood education in Ontario primary schools, which require two educators per classroom. High versus low transmission rates were also contrasted, keeping with the putative benefit of infection control measures within centers, yielding a total of 12 distinct scenarios. Simulations revealed that the 7:3 siblings together configuration demonstrated the lowest risk, whereas centres hosting classrooms with more children (15:2) experienced 3 to 5 times as many COVID-19 cases. Across scenarios, having less students per class and grouping siblings together almost always results in significantly lower peaks for number of active infected and infectious cases in the institution. Importantly, the total student-days lost to classroom closure were between 5 and 8 times higher in the 15:2 ratios than for 8:2 or 7:3. These results suggest that current proposals for childcare reopening could be enhanced for safety by considering lower ratios and sibling groupings.