Anthropic Launches Claude Science AI Workbench for Researchers
Anthropic has released Claude Science, an AI workbench designed to accelerate scientific discovery by integrating commonly used research tools and packages. The workbench provides auditable artifacts, flexible compute access, and specialized agents for various scientific domains, aiming to streamline complex research workflows. It is now available in beta for Claude Pro, Max, Team, and Enterprise users, offering enhanced reproducibility and computational management. This release allows scientists to analyze literature, execute multi-step research, and refine figures and manuscripts within a unified environment.
- →Claude Science AI Workbench for Scientists
- →Integrated Scientific Artifact Generation and Reproducibility
- →On-Demand Compute Management for Large Analyses
- →Domain-Specific Agent Capabilities and Customization
- →Beta Availability and User Feedback
Features (4) ›
- Claude Science AI Workbench for Scientists
Claude Science is an AI workbench that integrates research tools, produces auditable artifacts, and offers flexible compute access, aiming to accelerate scientific discovery and development in areas like healthcare. It brings fragmented tools into a single environment for literature analysis, multi-step research execution, and iterative refinement of figures and manuscripts, ensuring every output has an auditable history.
- Integrated Scientific Artifact Generation and Reproducibility
The workbench natively renders rich scientific artifacts including 3D protein structures and genome browser tracks, generating figures and manuscripts alongside the code that created them. Users can interact with the agent to edit figures and manuscripts in plain language, with all outputs including the exact code, environment, and message history for validation and reproduction.
- On-Demand Compute Management for Large Analyses
Claude Science handles the process of setting up, running, and retrieving results for large analyses by drafting plans, seeking resource approval, and submitting jobs to existing lab infrastructure like HPC clusters or Modal accounts. It scales analyses as needed and ensures large or sensitive datasets remain on the user's systems, with agents working within a context-holding session to load datasets only once.
- Domain-Specific Agent Capabilities and Customization
Specialist agents query and synthesize information across hundreds of specialized scientific sources, including connections to NVIDIA's BioNeMo Agent Toolkit for life sciences models. Claude Science can connect to existing lab models, datasets, and pipelines, saving them as reusable skills or connectors for future sessions, allowing access to proprietary data and tools within a single conversation.
Notes (2) ›
- Beta Availability and User Feedback
Claude Science is currently in beta and available for Claude Pro, Max, Team, and Enterprise users. Anthropic will continue to refine the platform based on user feedback collected during this beta period.
- Use Cases Demonstrated by Beta Testers
Researchers have used Claude Science in beta for tasks such as single-cell RNA sequencing analysis, CRISPR screen design, and protein structure prediction. Examples include Manifold Bio nominating targets for tissue-targeting medicines and the Allen Institute building a multi-agent 'computational review template' to expedite long-form reviews.
https://www.anthropic.com/news/claude-science-ai-workbench