NVIDIA’s open-source BioNeMo Agent Toolkit integrates with Anthropic’s newly launched Claude Science public beta. The integration gives life sciences researchers access to GPU-accelerated AI workflows for genomics, proteomics and drug discovery without manual tool configuration. NVIDIA Blog
Scientists can describe research tasks in natural language, and Claude Science orchestrates work through domain-specialized agents that call preconfigured, GPU-accelerated tools from the BioNeMo toolkit. Analyses that previously took hours now run in minutes, eliminating the need to manually configure models, endpoints or software environments. For example, a researcher can request a protein structure prediction or a chemical similarity search without writing code or provisioning GPU instances. NVIDIA Blog
BioNeMo Agent Toolkit Cuts Research Workflow Times in Claude Science Beta
The BioNeMo Agent Toolkit bundles purpose-built, GPU-accelerated tools for core life sciences workloads, preconfigured for immediate use by AI agents without manual software optimization. NVIDIA Blog
NVIDIA Parabricks, included in the toolkit, compresses end-to-end genomic analysis workflows from hours to minutes. This compression enables near-real-time genomic context integration into agent reasoning loops. NVIDIA Blog
RAPIDS-singlecell, developed by the scverse community and included in the toolkit, reduces a 1.3-million-cell preprocessing and clustering job from 52 minutes to 25 seconds. This speedup makes single-cell analysis feasible as part of iterative, real-time research rather than an offline batch process. NVIDIA Blog
For cheminformatics workloads, nvMolKit accelerates molecular similarity search and conformer generation by up to 3,000x. This speedup lets agents iterate across massive chemical libraries without manual bottlenecks in early-stage drug discovery. NVIDIA Blog
The toolkit also includes accelerated open biomolecular models including Evo 2, Boltz-2 and OpenFold3, packaged as NVIDIA NIM microservices for stable, production-ready inference endpoints. Agents can call these models via a single API, with no custom model deployment or endpoint management required. NVIDIA Blog
Open Design Enables Cross-Platform Agentic Research Workflows
The BioNeMo Agent Toolkit is open-source and harness-agnostic, meaning its prebuilt scientific skills work across agent frameworks and research platforms, not just Claude Science. The toolkit and all associated skills are available immediately via NVIDIA developer resources and public GitHub repositories. NVIDIA Blog
18 of the top 20 global pharmaceutical companies already use NVIDIA BioNeMo for AI-enabled research across drug discovery, molecular design and protein engineering. This integration marks a shift from generic AI assistants to domain-specific agentic systems built for the unique computational requirements of life sciences research, including high-throughput protein folding and single-cell analysis. NVIDIA Blog
Sample Cancer Inhibitor Workflow Demonstrates End-to-End Acceleration
NVIDIA outlined a sample end-to-end cancer research workflow to demonstrate the integration’s value for real-world use cases. A researcher starts with a known cancer-causing antigen mutation, then asks Claude Science to design potential inhibitors. The integrated system uses BioNeMo Agent Toolkit tools and NVIDIA NIM microservices to accelerate high-throughput inhibitor prediction, optimization and validation. NVIDIA Blog
As part of the Claude Science public beta, Anthropic is inviting researchers to submit feedback on additional domain specialists and integrations they need for their workflows. This feedback will inform future updates to both Claude Science and the BioNeMo Agent Toolkit skill library. NVIDIA Blog
