Our gut flora, the microbiome, plays an indispensable role in our health. Changes in the microbiome have been linked to an increasing list of diseases. Metagenomics is a technique that allows the sequencing of microbiomes directly from samples, giving valuable insight into the composition and functional potential of microbial populations. The analysis of metagenomic data is complex and depends on the availability of reference genomes. This work describes computational methods that allow the analysis of microbiomes with a lack of reference genomes by assembling genomes from the metagenomic data. We demonstrate how our methods can be used to infer the functional potential of a microbiome and how they allow us to link each function to the responsible species. We could predict changes in metabolites that were confirmed by targeted measurements. The mouse is the most used model for studying the impact of microbiota on its host. However, the species living in the mouse gut remain poorly characterized. By analyzing all publicly available metagenomes from the mouse gut, we created a comprehensive catalog of all bacterial species commonly living in the gut of laboratory mice. We assembled over 30'000 bacterial genomes, as well as the sequences from viruses and plasmids. Our catalog effectively answers the need for reference genomes for this microbiome. It allows efficient analysis of mouse gut metagenomes at the species and subspecies level. We discovered that mice and humans harbor a largely distinct set of species in their gastrointestinal tracts, an analysis which was hereto unfeasible.