Anthropic Launches Claude Science: AI Workbench for Researchers

Anthropic has officially launched Claude Science, a specialized AI workbench designed to streamline the research process for scientists. By integrating fragmented tools and databases into a single, cohesive environment, the platform aims to accelerate scientific discovery and the development of healthcare interventions. The app is available in beta for Claude Pro, Max, Team, and Enterprise users, providing a unified interface for literature analysis, multi-step research execution, and the creation of publication-ready manuscripts and figures.

A Unified Research Environment

Scientific research often requires navigating dozens of databases, bespoke data pipelines, and a roster of disparate tools such as Jupyter, R, and cluster terminals. Claude Science addresses this complexity by bringing these fragmented resources into one environment. Users interact with a generalist coordinating agent that has access to over 60 curated skills and connectors pre-configured for fields including genomics, single-cell analysis, proteomics, structural biology, and cheminformatics.
The platform emphasizes reproducibility by producing auditable artifacts. Every output includes the exact code, environment, and message history used to generate it, allowing researchers to validate results months later. Furthermore, a built-in reviewer agent monitors the process, checking citations and calculations to flag or correct errors in real time.

Scalable Compute and Domain Expertise

Claude Science manages computational resources by drafting plans and submitting jobs to the infrastructure a lab already uses, such as local machines, HPC clusters via SSH, or Modal accounts. Because the agents operate within a running session that maintains context in memory, large datasets do not need to be repeatedly reloaded. This approach allows sensitive data to remain on the lab’s own systems, with only the necessary context sent to Claude for analysis.
The workbench is also designed to synthesize information across hundreds of specialized sources, such as UniProt, PDB, Ensembl, and ChEMBL. By utilizing the NVIDIA BioNeMo Agent Toolkit, it connects natively to life sciences models like Evo 2, Boltz-2, and OpenFold3. Researchers can also integrate their own proprietary data and pipelines, saving them as reusable skills for future sessions.

Real-World Applications and Support

Early adopters have already utilized Claude Science to advance complex research projects. Manifold Bio used the platform to nominate tissue-targeting medicine candidates by assessing surface expression and safety criteria against internal data. Neuroscientist Jérôme Lecoq of the Allen Institute developed a multi-agent system to automate the writing of long-form literature reviews, reducing a process that previously took two years to a significantly faster workflow. Additionally, epidemiologist Stephen Francis at the UCSF Brain Tumor Center reported that the tool accelerated his molecular epidemiology studies on glioma by roughly ten times.
To support further innovation, Anthropic is offering up to 50 AI for Science project grants, providing up to $30,000 in credits, with Modal contributing up to $2,000 in additional compute for select projects. Applications for these projects are open through July 15, 2026.