Claude Science, an AI workbench for scientists, is now available
Anthropic News · 2026-06-30
Anthropic launches Claude Science in beta, an AI workbench integrating 60+ domain-specific scientific tools and a multi-agent architecture to let researchers conduct literature review, data analysis, and manuscript preparation in a single auditable environment.
Appears in
Extraction
Topics: ai-for-sciencemulti-agent-systemslife-sciencesresearch-productivityanthropic
Claims
- Claude Science integrates over 60 curated skills and connectors pre-configured for genomics, single-cell, proteomics, structural biology, and cheminformatics, accessible from a single coordinating agent.
- Every output produced by Claude Science includes the exact code, execution environment, and full message history used to generate it, enabling reproducibility and independent validation.
- Claude Science connects to NVIDIA's BioNeMo Agent Toolkit to access specialized life sciences models including Evo 2, Boltz-2, and OpenFold3.
- A reviewer agent automatically checks citations, traces numbers, and verifies that figures match their underlying code, self-correcting errors during the pipeline run.
- Early beta users report order-of-magnitude speedups: a UCSF epidemiologist completed germline analyses in one-tenth the previous time, and the Allen Institute reduced two-year review timelines to weeks.
Key quotes
Claude Science brings these fragmented tools into a single research environment where scientists can conduct all stages of their work.
Every output carries an auditable history of how it was made, so you can validate and reproduce the results.
Before Claude Science, it could take Lecoq's team as many as two years to write such a review. He now has about 10 reviews, many more than 100 pages, with citations that were checked over by reviewer agents.