Anthropic has launched Claude Science in public beta, an AI workbench designed for scientific research that now integrates NVIDIA's BioNeMo Agent Toolkit. The platform lets researchers conduct end-to-end computational workflows by conversing with digital agents in natural language rather than writing code manually.
Claude Science connects researchers directly to specialized AI agents that handle complex life sciences tasks. Instead of managing separate tools and scripts, scientists describe their research goals conversationally. The system interprets these requests and executes the necessary computational steps. This approach removes friction from routine research work, letting scientists focus on hypothesis design and interpretation.
NVIDIA's BioNeMo toolkit specifically accelerates molecular and genomic analysis within Claude Science. BioNeMo includes pretrained models for protein structure prediction, molecular property analysis, and sequence processing. By embedding this specialized toolkit into Claude Science's conversational interface, researchers gain access to production-grade bioinformatics capabilities without infrastructure setup or deep machine learning expertise.
The timing reflects growing demand for accessible AI in wet labs and computational biology departments. Academic labs and biotech startups often lack dedicated machine learning engineers or sufficient computing budgets to build custom pipelines. Claude Science addresses this by packaging research automation into a chat interface that works like a domain-expert collaborator.
Anthropic positions Claude Science as a research-focused alternative to general-purpose AI assistants. The platform emphasizes reproducibility and scientific rigor over speed. Researchers can document their conversational workflows, making them auditable and shareable with collaborators.
The integration with BioNeMo marks a shift toward specialized AI ecosystems for specific research domains. Rather than hoping general models understand biology, NVIDIA and Anthropic jointly embedded domain knowledge into the toolkit itself. This allows Claude Science to handle sequence alignment, protein folding, and molecular docking with greater accuracy than generic language models would provide.
The public beta