AI-native tools for biology, chemistry, and materials science.
DarwinOmics gives AI agents a single, verified toolset for computational science: 7,666 tools, models, and pipelines in build, each with a machine-readable interface and a public record. The platform is in active development; release access is currently by inquiry.
platform overview
in build
What DarwinOmics does
One platform across three domains.
A unified catalog
7,666 tools, models, visualizations, and pipelines across biology, chemistry, and materials science, each with a public record page.
Machine-readable interfaces
Every entry ships with a machine-readable contract, so an AI agent can select and run tools without custom glue code.
Verified against references
Where a reference implementation exists, module output is checked against it under pinned, declared conditions.
Grounded in domain knowledge
A knowledge graph and a graph neural network sit under the corpus, supporting genomics, drug discovery, and materials queries.
Three domains, one platform
The same interface, in every domain.
Biology
Reads, variants, structures, and cells: the working tooling of molecular biology, rebuilt for machine operation.
2,780 modules in build · 2,066 cataloged
Chemistry
Docking, dynamics, reactions, and formulations, operated through the same interface as the rest of the platform.
2,486 modules in build · 1,147 cataloged
Materials science
Crystals, battery materials, and property prediction, cataloged to the same standard as every other module.
2,400 modules in build · 122 cataloged
Release standards
A module releases when it clears the bar.
Checked against the original
Where a reference implementation exists, module output is verified against what it actually produces, under pinned and declared conditions.
The record travels with the tool
Every entry carries a provenance record: what it was compared against, under which conditions, and what was produced.
One bar for all 7,666
There is no fast lane and no showcase tier. A module releases when it clears the same standard as every other module.
Model-agnostic by design
Any AI agent can operate the platform.
Tools communicate through a machine-readable manifest rather than a proprietary API or runtime. The model you already use is the interface.
Reads the manifest once and operates the corpus the same way it operates any other instrument.
The same tools and the same surface, with a different agent in control.
Local or hosted, small or large. Anything that can call a function can run an aligner, fold a protein, or screen a battery material.
Built on this corpus and trained to operate it. They will be released openly.
Release channels
Three ways to run it.
The full corpus, packaged for desktop systems, including builds that run on modest hardware. No AI attached: the scientific tools on their own.
Distributed through the Microsoft Store and Apple App Store.
The corpus with its AI interfaces enabled: operate it with Claude or Codex, or run the DarwinOmics agents on your own machines.
Knowledge-graph reference access ships in the same channel.
The full platform installed within your institution, configured for your security and compliance requirements.
Your data remains within your infrastructure.
In development, with a public record.
The catalog documents the platform as it is built. For access timelines, packaging, or partnership inquiries, contact the team.