Anthropic announced Claude Science, an AI workbench designed for scientists. Claude Science is an app that integrates common research tools and packages, produces auditable artifacts, and provides flexible access to computing resources.
The platform is intended to help scientists conduct research across a single environment rather than moving between databases, file formats, tools, notebooks, terminals, and compute systems.
Anthropic said Claude Science supports literature analysis, multistep research execution, figure and manuscript refinement, and publication-ready scientific artifacts.
Every output includes an auditable history showing how it was made, helping researchers validate and reproduce results.
Claude Science is available in beta for Claude Pro, Max, Team, and Enterprise users.
The app can be used locally on macOS or Linux, or on a remote machine over SSH or with an HPC login node.
Users interact with a generalist coordinating agent that has access to more than 60 curated skills and connectors pre-configured for genomics, single-cell analysis, proteomics, structural biology, cheminformatics, and other scientific domains.
The system can also spin up specialist agents and interact with specialist agents created by users.
A reviewer agent checks citations and calculations, flags errors, and helps correct issues.
Claude Science can generate scientific figures and manuscripts alongside the code that created them.
The platform natively renders scientific artifacts such as 3D protein structures, genome browser tracks, and chemical structures.
When Claude Science generates a figure, it includes the exact code and environment used to produce it, along with a plain-language explanation and the full message history.
Researchers can request figure edits in plain language, such as removing gridlines or changing an axis to a log scale, and the agent edits its own code.
Claude Science also manages compute resources for large analyses, such as protein folding or genomics pipelines over large datasets.
The app can draft a plan, ask before using new resources, and let users review or revoke decisions before submitting jobs to the computing resources a lab already uses, including HPC clusters over SSH or Modal compute on demand.
Anthropic said the platform can scale analyses from a single GPU to hundreds as needed.
Because agents operate inside a running session with context in memory, large datasets only need to be loaded once.
Claude Science runs on the user’s infrastructure, which means large or sensitive datasets can stay on existing systems while only the context needed for each step is sent to Claude.
Claude Science can also query and synthesize across specialized scientific sources and databases.
In biology, this can include resources such as UniProt, PDB, Ensembl, Reactome, ClinVar, ChEMBL, GEO, journals, preprint servers, and domain-specific open models.
The platform also uses skills in NVIDIA’s BioNeMo Agent Toolkit to connect to life sciences models and libraries in BioNeMo, including Evo 2, Boltz-2, and OpenFold3.
Claude Science can connect to trusted lab models, datasets, and pipelines, and save pipelines as reusable skills for future sessions.
Anthropic said researchers have used Claude Science in beta for tasks including single-cell RNA sequencing analysis, CRISPR screen design, protein structure prediction, cheminformatics, and long-form scientific reviews.
Manifold Bio used Claude Science to nominate targets for tissue-targeting medicines by assessing surface expression, trafficking, and safety while incorporating context from prior programs.
Jérôme Lecoq, a neuroscientist at the Allen Institute, used Claude Science to build a multi-agent computational review template with about 20 custom skills for writing long-form reviews.
Stephen Francis, an associate professor and epidemiologist at the UCSF Brain Tumor Center, used Claude Science to support studies on the molecular epidemiology of glioma.
Anthropic is also supporting up to 50 Claude Science AI for Science projects with up to $30,000 in credits, while Modal will provide up to $2,000 in compute for select projects.
Applications are open through July 15, 2026, with award notifications expected by July 31.
The projects will run from September 1 to December 1, 2026.
