Anthropic has launched Claude Science, an AI workbench designed to help researchers manage literature reviews, data analysis, computing workflows and scientific writing in one environment.

The app can produce figures, manuscripts and other scientific artifacts alongside the code and computing environment used to create them. Each output includes a record of its inputs and message history, allowing researchers to inspect, reproduce and revise the results.

Claude Science includes more than 60 skills and connectors covering genomics, single cell analysis, proteomics, structural biology and cheminformatics. It can query sources including UniProt, PDB, Ensembl, ClinVar, ChEMBL and GEO, while connecting to life sciences models and libraries through Nvidia’s BioNeMo Agent Toolkit.

Researchers interact with a coordinating agent that can assign work to specialist agents. A separate reviewer agent checks citations, calculations and figures, flagging results that cannot be traced to their underlying data or code.

The app can also prepare and submit computing jobs to a laboratory’s existing infrastructure, including remote machines and high performance computing clusters accessed through SSH. Users can review or revoke decisions before Claude connects to additional resources.

Anthropic said large or sensitive datasets can remain on a researcher’s laptop, Linux machine or laboratory cluster, with only the context required for each stage of the analysis sent to Claude. The platform can also scale jobs from one graphics processor to hundreds through Modal.

Claude Science is available in beta on macOS and Linux for Pro, Max, Team and Enterprise customers. Team and Enterprise administrators must enable access before members can use the app.

Anthropic will support as many as 50 research projects with up to $30,000 in Claude credits each, while Modal will provide up to $2,000 in computing credits for selected projects. Applications close July 15, with projects scheduled to run from September through December.