Anthropic Integrates NVIDIA BioNeMo into Claude Science Workbench

The integration bridges the gap between human research reasoning and machine-accelerated execution. By utilizing BioNeMo as a foundation, Claude Science translates plain-text requests into programmatic actions, effectively managing predictive models and network endpoints without requiring manual software configuration. Currently, 18 of the top 20 global pharmaceutical companies already rely on BioNeMo in their production environments, signaling widespread industry adoption of the underlying technology.

Scaling Computational Research

Beyond simple automation, the toolkit provides agents with specialized skills, such as genomic sequencing, protein structure prediction, and molecular binder design. By packaging these capabilities as callable NIM microservices, the system ensures that research pipelines—ranging from single-cell analysis to cheminformatics—operate at high speed. For instance, the RAPIDS-singlecell tool can reduce processing times for 1.3 million cells from 52 minutes to just 25 seconds, enabling agents to incorporate complex data into their reasoning loops in near real-time.

Researchers gain access to advanced open models including Evo 2, Boltz-2, and OpenFold3, all optimized through NVIDIA software libraries. Because the BioNeMo toolkit remains model-agnostic, these scientific skills function consistently across various enterprise frameworks. Engineering teams can access the toolkit via GitHub, with Anthropic currently soliciting feedback during the public beta to refine future domain-specific integrations.