Anthropic launches Claude Science for research workflows

Anthropic on Tuesday launched Claude Science, a beta workbench for scientists that puts literature search, data analysis, code and compute access inside one Claude environment.

Claude Science is available on Pro, Max, Team and Enterprise plans, the company said, and ships with more than 60 scientific databases, tools and curated connectors used in research settings. The release moves Anthropic further beyond the general chatbot market and into software for paying professional teams. It is a product launch rather than a model launch.

There is a commercial point underneath the lab pitch. Anthropic is packaging Claude as a workspace where researchers can search papers, run code, examine datasets and keep a record of the steps taken. Bloomberg reported that the product fits a push by frontier AI companies to sell higher-value workflow software instead of competing only on model benchmark scores.

Researchers already have plenty of windows open.

In practice, a team might move between papers, code notebooks, storage buckets, lab software and cloud compute before a result is ready to share. Anthropic said Claude Science is meant to bring those jobs into one interface while producing auditable artefacts that can be checked or reproduced later. That traceability claim matters in labs, where a useful answer from an AI assistant is not enough if a team cannot see where it came from, rerun the work or explain the method to a reviewer.

In its announcement, Anthropic said: “Claude Science is an app that integrates the tools and packages that researchers most commonly use, produces auditable artifacts, and provides flexible access to computing resources.” The company also pointed to prebuilt skills and flexible compute access, two details aimed at the setup work that stops many general AI trials from moving beyond summaries. Allen Institute neuroscientist Jérôme Lecoq was named as a beta user, signalling an early pitch to life sciences teams.

To seed adoption, Anthropic said it would support 50 Claude Science AI for Science projects with up to $30,000 in credits each. That subsidy gives labs and universities room to try the platform before habits form around rival tools from Google, OpenAI or internal software teams.

Bloomberg’s report said Anthropic sees the product as a way to reduce tedious data-handling and computing work for scientists. The commercial test is whether frontier AI companies can turn those use cases into durable software revenue. Model performance still draws the headlines; specialist products for researchers, developers and enterprise teams are more likely to decide which vendors win recurring budgets.

No Australia-specific pricing, research partnerships or local customers were announced on Tuesday. Even so, the release matters for Australian universities, medical institutes and corporate R&D teams already testing generative AI, because vendors are now selling integrated workbenches rather than chat interfaces alone.

For those institutions, the next questions are closer to procurement than hype: which environment keeps enough of the research trail, and which one can sit inside day-to-day scientific work without making governance harder. Claude Science is Anthropic’s bid to make those questions about its software, not just its model.