NVIDIA integrates BioNeMo toolkit into Claude Science
NVIDIA says it partnered with Anthropic to wire its BioNeMo Agent Toolkit into Claude Science, letting agents call GPU-hosted protein-folding microservices directly — a rare named integration between the two companies in AI-for-science tooling.
NVIDIA said it worked with Anthropic to integrate the BioNeMo Agent Toolkit into Claude Science, Anthropic's workbench for scientific research, letting agents discover, launch, and call BioNeMo NIM microservices directly. The disclosure came in a tutorial post on NVIDIA's developer blog rather than a formal announcement from either company.
The toolkit packages NVIDIA's BioNeMo life-sciences models, libraries, and workflows into agent-callable skills for biology, chemistry, genomics, and drug discovery, and is built to run with any agent framework. NVIDIA claims that on internal benchmarks BioNeMo skills raise task correctness from 60% to 100% and roughly double token efficiency. Those benchmarks are not published, and the post gives no detail on task set, baseline agent, or evaluation method.
The walkthrough runs protein structure prediction on a two-protein system from the fungus Paracoccidioides lutzii — the nuclear-pore protein Seh1 (C1GY11) and an uncharacterized proposed partner (C1HCX1) — using the msa-search, OpenFold3, and Boltz-2 NIMs, with the example motivated by Figure 4e of Han, Tsenkov, Venanzi et al., "AlphaFold Database expands to proteome-scale quaternary structures" (bioRxiv, DOI 10.64898/2026.03.27.714458). The search returned 202 sequences for each protein.
Setup is heavy: an NVIDIA L40S or H100 GPU and roughly 700 GB of storage, including the ~490 GB UniRef30-only msa-search profile in place of the full 1.4 TB set, plus 30–40 GB of Boltz-2 and OpenFold3 containers. Endpoints are configured in Claude Science under Customize > Compute, with skills imported from GitHub and an NVIDIA API key.