Overview
- Anthropic launched the Claude Science beta, which it announced June 30, and made the workbench available to Pro, Max, Team and Enterprise users on macOS and Linux with options to run locally, over SSH, or on HPC login nodes.
- Claude Science captures auditable artifacts for reproducibility by bundling code, execution environment, plain‑language descriptions, and message history with generated figures and by offering a reviewer agent that checks citations and calculations.
- The workbench links to NVIDIA's BioNeMo Agent Toolkit skills and life‑sciences models and libraries such as Evo 2, Boltz‑2, and OpenFold3 to run domain‑specific analyses and experimental workflows.
- On July 2 Anthropic expanded Claude Desktop enterprise deployment so organizations can run inference in AWS, Google Cloud, or Microsoft Foundry while using per‑user SSO, mobile device management templates, offline installers, separate policy keys per product surface, and a Microsoft 365 connector that supports Entra tenant allowlisting and beta GCC High and DoD endpoints.
- These products reflect a wider shift from chat assistants to agentic, workflow‑capable platforms that let researchers trace and reproduce results and give IT teams the identity, policy and data‑residency controls needed for regulated labs and large enterprises, and Anthropic has also said it plans to pursue drug discovery work using these tools though it has given few public details.