Anthropic launches Claude Science, an AI workbench for researchers

Anthropic has announced the release of Claude Science, a research environment that consolidates multiple scientific databases, tools, and visualization capabilities into a single platform for scientists and researchers.

Anthropic announced the product at its “The Briefing: AI for Science” event on June 30, where the company explained Claude Science was not a new AI model. The product runs on the Claude models already available to subscribers and will function as a workbench where researchers can assess scientific literature and run analyses without switching between resources.

Claude Science available in beta, is open to all Claude Pro, Max, Team, and Enterprise subscribers, according to Anthropic’s announcement.

Claude Science a hub for research

A central AI assistant handles research tasks and connects to a list of over 60 scientific databases preconfigured for different research areas like genomics, proteomics, single-cell analysis, and cheminformatics, according to the statement. This central assistant can create sub-agents for specialized tasks and even delegate work to custom agents built by researchers for their own particular interests.

A separate reviewer agent will checks citations and calculations, and flag errors before results are put into a manuscript. This step will help to address the fact-checking and verification problem that could creep into AI-assisted papers; however, Anthropic cautioned that the checker still relies on the same underlying model, not an independent source for verification.

Every output produced by the workbench carries what Anthropic calls an “auditable history”. This entails every step that led to the output, including the code that produced a figure, the environment the code ran in, an easy-to-understand description of the methodology, and the full message log.

Researchers can fork a session at any point to compare two analytical approaches without losing the original thread of events.

Anthropic’s early adopters and partnerships

Jérôme Lecoq, a neuroscientist at the Allen Institute, built a multi-agent flow within the platform that helped with reading thousands of papers and extracting important claims and quantitative findings.

Stephen Francis’s group at the UCSF Brain Tumor Center used the tool to compress germline analysis of glioma into a fraction of the time previously used for the same analysis, TechCrunch reported.

Anthropic also named Novo Nordisk and the Allen Institute as customer case studies, and is offering up to $30,000 in credits for as many as 50 research projects. Applications for this grant will close July 15, 2026, and recipients will be announced by July 31, 2026.

Scientific AI sees development due to competition

Direct competitor to Anthropic, OpenAI, released GPT-Rosalind in April 2026, which was an AI model specifically tuned for biological reasoning but limited to qualified enterprise customers in the U.S. only after a safety review. Google DeepMind’s foundational science models, like AlphaFold and AlphaGenome, have been bundled together with more than 30 life science databases through its Gemini for Science platform.

Anthropic is looking at satisfying its already owned customers with inclusive subscription access alongside easy integration instead of creating a new or specialized model. The company has filed a confidential IPO with the SEC, and its moves to expand into vertical products like Claude Science fit a pattern of building more revenue streams ahead of its public listing, according to Yahoo Finance.