The platform combines a coordinating AI agent with more than 60 preconfigured skills and connectors covering fields including genomics, single-cell biology, proteomics, structural biology, and cheminformatics. Users can also create specialist agents tailored to their own workflows, while a separate reviewer agent checks citations, calculations, and generated outputs for potential errors.
Anthropic said Claude Science can manage computational workloads across existing research infrastructure, including local machines, high-performance computing clusters accessed over SSH, and on-demand GPU resources through Modal. Researchers retain control over the process, with the application requesting approval before using new computing resources or submitting jobs.
According to the company, the application keeps large datasets on researchers’ existing infrastructure, sending only the context required for each stage of an analysis to Claude. Scientists can also branch ongoing projects to compare different approaches while preserving previous work.
Claude Science includes native support for scientific content such as protein structures, genome browser tracks, chemical structures, and other research artifacts. Each generated figure is accompanied by the code, execution environment, and a plain-language explanation describing how it was produced, allowing researchers to review and reproduce results later. Figures and manuscripts can also be revised using natural language instructions, with the application updating its own underlying code to reflect requested changes.
Anthropic said the platform integrates with NVIDIA’s BioNeMo Agent Toolkit, providing access to life sciences models and libraries including Evo 2, Boltz-2, and OpenFold3. Scientists can also connect Claude Science to their own datasets, pipelines, and laboratory tools, saving those workflows as reusable skills for future sessions.
The company highlighted several early users that tested Claude Science during its beta program. Manifold Bio said it used the platform to rank drug targets by evaluating factors such as surface expression, trafficking, and safety using its proprietary data. The company said Claude Science could complete the workflow end-to-end rather than functioning solely as a coding assistant.
Jérôme Lecoq, a neuroscientist at the Allen Institute, built a multi-agent workflow with approximately 20 custom skills for writing scientific review papers. Anthropic said the system analyzes thousands of research papers, extracts key findings into an evidence database, and uses specialized agents to draft sections and generate quantitative figures, while reviewer agents verify citations and factual accuracy. According to the company, Lecoq’s team has produced around 10 reviews, many exceeding 100 pages, after a process that previously took as long as two years.
Stephen Francis, an associate professor and epidemiologist at the UCSF Brain Tumor Center, has used Claude Science to support molecular epidemiology research on glioma. Anthropic said Francis independently validated the application’s results and reported that it reduced analysis time to roughly one-tenth of previous workflows while maintaining robust results.
Alongside the product launch, Anthropic announced a discounted Team plan for scientific laboratories at academic institutions and nonprofit research organizations. The company also said it will support up to 50 AI for Science projects with as much as $30,000 in Claude Science credits, while Modal will provide up to $2,000 in computing resources for selected projects. Applications are open through July 15, 2026, with projects scheduled to run from September 1 through December 1, 2026.
During its “The Briefing: AI for Science” event, Anthropic also said it plans to develop its own drugs focused on neglected diseases. The company did not provide additional details about disease targets, partnerships, or development plans.
This analysis is based on reporting from Anthropic.
Image courtesy of Anthropic.
This article was generated with AI assistance and reviewed for accuracy and quality.
