The living corpus.

Biology is where this company started, and it shows: 2,780 modules in build, from raw reads to drug targets. Every aligner, caller, folder, and screener is rewritten to be driven by a machine and checked against the tool it succeeds.

2,780
Modules in build
2,066
Corpus entries
33
Families on the shelf
3 builds
One aligner at a time

What 2,066 records hold

The corpus, by species of entry.

A catalog entry is a promise with a page. A native tool is the promise kept in Rust. Models, visualizers, and pipelines each get their own shelf; every row below links into the catalog.

Catalog entries
Tools absorbed into the corpus and given records, contracts, and a shelf.
1,575entries
Native tools
Written from scratch in Rust where no reference existed to follow.
149entries
Models
Trained weights with their evaluation records attached.
108entries
Visualizers
Figure engines and browsers for what the corpus produces.
108entries
Pipelines
Multi-step runs that chain the corpus into one sentence.
106entries
BioNeMo-lineage models
Foundation-class models for sequence and structure.
20entries

The flagship, in three sizes

One aligner, three builds.

The most famous aligner in biology has an infamous appetite. Ours ships three ways: full speed, balanced, and a build that fits inside 500 megabytes of memory. The third one exists for the laptop under a grant budget, the field station with one outlet, and everyone whose machine was never the point.

All three are free at launch, with no AI attached, and they verify against the original's output.

aligner buildsin build
buildfast
forworkstations with room to run
buildstandard
forthe everyday server
buildlow-memory
for500 MB of RAM. Yes, MB
Same output, same rigor, three footprints. A tool should fit the machine that has the question.

Under the tools

The grounding layer.

Tools answer questions. Grounding decides which questions are worth asking. A knowledge graph for the human and plant genomes and a graph neural network for genomics, drug discovery, and repurposing sit under the agents, reading alongside your data.

Reference and citation access for researchers: limited and free in the coming months, wider by subscription.

Human genome
one of the largest graphs assembled
Plant genome
the quiet half of the field
Drug discovery
targets, compounds, repurposing
Graph neural network
reads the graph natively, no export
Chem + materials
grounding layer, next in the build
Rare disease
where tangential analysis matters most

The shelf

Thirty-three families, one address.

Free
the whole corpus, no AI

Every bioinformatics entry, packaged for ordinary machines and stores you already use. The low-memory STAR build ships here first because that is where it matters.

coming soon
Subscription
the same corpus, driven

Claude, Codex, or our own agents holding the wheel. Reference access to the knowledge graph lands in the same apps.

coming soon
Enterprise
in your walls

The substrate on your infrastructure, your data behind your gates, your compute doing the carrying.

coming soon

Biology keeps its own books now.

2,780 modules with pages you can read today and binaries you can run someday soon. Nothing ships until it proves itself.