The microbiome field is rich in data and short on shared scaffolding. Most labs rebuild the same pipelines, annotations, and embeddings from scratch. The work rarely makes it into a paper, so it rarely makes it out of the lab.

We're releasing what we'd want if we were starting fresh. Open weights so labs can fine-tune on their own cohorts. Open code so results are reproducible without a licence. Open data, where upstream terms allow it, so benchmarks compound across groups instead of fragmenting.

Foundation models have transformed structural biology and genomics, but microbiome science hasn't had its moment yet. We think the timing is right. Our work builds on the open-source tools and public datasets the community has spent years creating, and this is our contribution back.