Anthropic launches Claude Science AI workbench for researchers
Anthropic on Tuesday unveiled Claude Science , a new AI workbench built around the idea that researchers should not have to shuttle between disconnected databases, pipelines, and analysis tools to get their work done. The product launched in beta for users on Pro, Max, Team, and Enterprise subscription plans.
The workbench connects to more than 60 scientific databases and comes with prebuilt toolkits for fields including genomics, proteomics, structural biology, and cheminformatics, the company said. A primary agent oversees the overall project, either spawning specialized sub-agents to divide the workload or routing tasks to custom agents that individual researchers have created for their own use cases. Before any output reaches publication, a dedicated reviewer agent independently audits citations and calculations, catching and correcting mistakes along the way. Anthropic said Claude Science runs on existing Claude models, including Opus 4.8, and is not a new or specialized model.
Scientists can run the workbench locally on macOS or Linux machines, on a remote server over SSH, or through a high-performance computing login node. Because the software runs on a lab's own infrastructure, Anthropic said large or sensitive datasets do not need to leave existing systems. For larger computational jobs — protein folding or a genomics pipeline across a large dataset, for example — the workbench can draft and submit the job to the lab's own HPC cluster or to Modal, a cloud compute service, scaling from a single GPU to hundreds as needed, the company said.
Anthropic said every figure the workbench generates is bundled with a reproducibility package — the underlying code, the computational environment, a plain-language explanation of the methodology, and the complete message history — so that other researchers can verify and replicate the work. Researchers can also describe desired changes to a figure in ordinary language, and the agent will revise its own code to carry them out.
Early users reported concrete time savings. Jérôme Lecoq of the Allen Institute described using Claude Science to construct an automated pipeline for producing scientific literature reviews; his team can now complete reviews exceeding 100 pages — work that previously stretched across as long as two years — with reviewer agents handling citation verification. At the UCSF Brain Tumor Center, associate professor and epidemiologist Stephen Francis reported that comprehensive germline analysis of glioma that once took his group substantial time could now be completed in roughly a tenth of that span, with independently validated results.
Anthropic's push into scientific workflow tools follows its partnership with Bristol-Myers Squibb , which gave the drugmaker's workforce access to Claude for research and drug development. The company launched Claude for Life Sciences in October 2025, and Claude Science represents a dedicated application environment built on top of that foundation.
Alongside Claude Science, Anthropic said it has started an internal drug discovery program focused on diseases that traditional pharmaceutical companies would not consider commercially attractive targets, according to CNBC . According to Eric Kauderer-Abrams, Anthropic's head of life sciences, running its own drug programs gives the company the hands-on experience it believes is necessary to credibly help life sciences customers rethink their own development workflows.
To encourage early adoption among researchers, Anthropic is offering grants of as much as $30,000 in credits to up to 50 qualifying Claude Science projects, with the application window closing July 15, 2026; selected projects are scheduled to run between September 1 and December 1.
