Anthropic has launched Claude Science, a new artificial intelligence workbench aimed at streamlining scientific research by bringing literature analysis, computational workflows and research documentation into a single platform.
Announced on Tuesday, Claude Science is designed to help researchers conduct complex, multi-step scientific investigations while producing fully auditable outputs. The application combines more than 60 pre-configured skills and connectors tailored for disciplines including genomics, single-cell analysis, proteomics, structural biology and cheminformatics.
Available for macOS and Linux, the platform can connect securely to remote computing environments via SSH or high-performance computing (HPC) login nodes. It enables researchers to generate manuscripts, figures and the underlying code simultaneously, while offering native visualisation of 3D protein structures, genome browser tracks and chemical structures.
Anthropic said the workbench can manage computational resources by drafting analysis plans and submitting jobs to existing HPC clusters or Modal cloud infrastructure. The system is capable of scaling workloads from a single graphics processing unit (GPU) to hundreds, depending on the requirements of a project.
To improve research transparency, Claude Science incorporates a reviewer agent that checks outputs for incorrect citations, figures that are inconsistent with the underlying code, and numerical results that cannot be traced back to their source.
The platform also integrates skills from NVIDIA’s BioNeMo Agent Toolkit, providing access to life sciences models including Evo 2, Boltz-2 and OpenFold3.
Early users have already applied the technology across a range of research settings. Biotechnology company Manifold Bio used Claude Science to identify potential experimental targets by evaluating tissue-specific surface expression, protein trafficking and safety profiles.
At the Allen Institute, Jérôme Lecoq developed a multi-agent computational review workflow capable of analysing thousands of scientific papers and producing narrative literature reviews supported by quantitative cross-study figures.
Meanwhile, Stephen Francis at the UCSF Brain Tumor Center employed the platform for molecular epidemiology studies on glioma, reducing the time required to complete comprehensive germline analyses to around one-tenth of previous workflows.
Claude Science is currently available in beta for Claude Pro, Max, Team and Enterprise subscribers.
Anthropic has also announced a new AI for Science programme that will support up to 50 research projects with credits worth up to 30,000 US dollars each. Applications are open until 15 July 2026, with successful applicants expected to be notified by 31 July 2026.
