Anthropic launches Claude Science workbench for researchers
TL;DR:
- Anthropic has launched Claude Science, an AI workbench that unifies the databases, tools and compute researchers juggle across a single environment.
- A coordinating agent draws on more than 60 curated skills for genomics, proteomics, structural biology and more, while a reviewer agent checks citations and calculations.
- It is available in beta for Pro, Max, Team and Enterprise users, with Anthropic offering up to $30,000 in credits for selected AI-for-science projects.
Anthropic has made its biggest move yet into research, releasing an app aimed at the tedium that surrounds actual science. Claude Science brings scattered resources — PubMed, Jupyter, R, cluster terminals and specialist databases — into one workspace where, the company says, scientists can run analyses, generate figures and refine manuscripts towards publication.
Reproducibility as the selling point
The pitch leans on auditability. Every figure ships with the exact code and computing environment that produced it, a plain-language account of how it was made, and the full message history — so results can be validated or reproduced months later. A reviewer agent inspects outputs as pipelines run, flagging untraceable numbers and figures that do not match their code. Sensitive datasets can stay on a lab’s own infrastructure, with only the context needed for each step sent to Claude.
Early users point to sharp time savings. A UCSF epidemiologist reported completing germline analyses in roughly a tenth of the previous time and independently validating the results, while an Allen Institute neuroscientist built a multi-agent pipeline that compresses two-year literature reviews.
The launch places Anthropic squarely in a widening life-sciences contest. It arrives the same day OpenAI unveiled GeneBench-Pro, a benchmark probing AI judgment in computational biology — evidence both labs see drug discovery and genomics as the next proving ground. For the UK’s research-heavy universities and life-sciences firms, tools that keep proprietary data in-house while accelerating analysis could prove particularly attractive.
Looking forward
Anthropic is funding up to 50 AI-for-science projects, with applications open until 15 July and work running from September. The open question is trust: reproducible artifacts and reviewer agents address the reliability concerns that have kept many researchers wary, but adoption will hinge on whether independent validation keeps confirming the outputs.