The beta consolidates researchers' scattered tooling under a coordinating agent and, Anthropic says, keeps sensitive data on a lab's own infrastructure.
Anthropic has released Claude Science, an AI workbench aimed at scientists, and made it available this month in beta. Per Anthropic's announcement, reported by STAT, the app opens to Claude Pro, Max, Team, and Enterprise subscribers.
The tool folds the fragmented stack researchers juggle — PubMed, Jupyter, R, cluster terminals, and dozens of databases — into a single environment where every stage of a project can run. According to Anthropic's announcement, users work through a generalist coordinating agent with more than 60 curated skills and connectors preconfigured for fields including genomics, single-cell biology, proteomics, structural biology, and cheminformatics.
What sets it apart is where it lives: Claude Science runs on a lab's own infrastructure — a laptop, Linux box, or HPC login node — so, Anthropic says, bulky or confidential datasets can stay put on the machines that already hold them, with only the context each analysis step requires sent to Claude. It can submit computing jobs to the lab's HPC cluster over SSH or to a user's Modal account for on-demand compute, Anthropic's announcement says.
Anthropic says outputs keep a traceable record of their creation, which users can review when checking or reproducing results. For a generated figure specifically, that record extends to the exact code and computing environment behind it, a plain-language explanation of how it came together, and the complete message history. As a pipeline runs, Anthropic's announcement adds, a reviewer agent audits each output on the fly — flagging citations that don't check out, figures whose numbers trace back to no source, and charts that stray from the code that produced them — then steers the run back on track.
The workbench also leans on NVIDIA's BioNeMo Agent Toolkit, using its skills to reach life-sciences models and libraries including Evo 2, Boltz-2, and OpenFold3, per Anthropic's announcement.
STAT framed the low-key unveiling more narrowly — as a tool that bends Anthropic's language model toward the everyday work of biopharma research labs. It lands, by STAT's account, in a year the outlet describes as one in which artificial intelligence has come to be widely viewed as a societal threat.