It is widely believed that Web search engines require immense resources to operate, making it impossible for small communities to build alternatives to the dominant players. The PeARS project (https://pearsproject.org/) aims at changing the status quo by providing open-source search tools that can be used by anyone, anywhere. To achieve this, it combines algorithms that run on entry-level hardware, using both traditional and modern machine learning techniques. This talk will show how PeARS can easily be deployed to provide community-based search solutions: for instance for underrepresented groups or minority languages. We will see how the system can be trained and populated and how it can be tailored to the specific community it is addressed to. We will also give a glimpse into cross-instance search and show how to search the entire PeARS network from a local instance.