Key Takeaways
- Anthropic launched Claude Science, an integrated platform for researchers, currently in beta for paid users.
- The app automates research workflows with AI agents managing tasks and ensuring citation accuracy.
- Claude Science supports various scientific databases and runs on existing lab infrastructure, safeguarding sensitive data.
Claude Science, developed by Anthropic, is a newly launched application aimed at simplifying scientific research. Unveiled on June 30, 2026, the platform consolidates various research tools, including databases, coding systems, and computational resources, into a single user-friendly workspace. Currently in beta, it is available to users subscribed to Claude.
One of the standout features of Claude Science is its ability to display scientific data—such as proteins, structures, and molecules—while ensuring that every result can be traced back to its original code. This commitment to transparency is an essential response to concerns about reproducibility in scientific research. The platform is built on Anthropic’s existing models, incorporating over 60 databases and toolkits tailored for diverse fields like genomics, proteomics, and structural biology.
Claude Science distinguishes itself by employing a two-tier agent system: a coordinating agent that facilitates tasks by assigning them to specialized sub-agents, and a reviewing agent that verifies citations and calculations throughout the research process. This design helps address the persistent issue of inaccurate citations often found in AI-assisted publications.
The application also emphasizes user control, allowing researchers to communicate tasks in plain language. The coordinating agent can generate other specialized agents to handle specific assignments while maintaining the flexibility for users to evaluate or cancel tasks before submission. Additionally, Claude Science can leverage a lab’s existing computational tools, scaling seamlessly based on the resources available.
In terms of security, Claude Science runs on a lab’s own machines, which means large datasets remain protected within their systems. The platform utilizes NVIDIA’s BioNeMo Agent Toolkit, which connects to prevalent life science models, enhancing its functionality while preserving data confidentiality.
Several early adopters of Claude Science have praised its capabilities. For example, Manifold Bio, a company specializing in targeted medicines, used the app for its latest research and found that it streamlined the entire task. Neuroscientist Jérôme Lecoq from the Allen Institute developed a pipeline with custom skills that drastically reduced the time spent generating literature reviews from two years to approximately three months. Similarly, Stephen Francis, an epidemiologist at UCSF, utilized the app to expedite glioma studies and validated results efficiently.
The development of Claude Science is part of a broader strategy by Anthropic to capture a dominant share of the life sciences sector. This aligns with their previous launch of Claude for Life Sciences in October 2025. By making the app broadly accessible, Anthropic aims to position itself as a leading tool for scientific research, mirroring successful models used in other industries, including law.
In addition to the app, Anthropic is initiating its own drug discovery programs, focusing on diseases typically overlooked by the pharmaceutical industry. This proactive approach allows the company to gain hands-on experience in lab environments, enhancing its relevance to future pharma clients.
Claude Science is currently available in beta for macOS and Linux users, requiring administrative activation for Team and Enterprise plans. To support research initiatives, Anthropic offers grants of up to $30,000 for 50 projects, along with additional compute credits from Modal. This program is aimed at fostering interdisciplinary collaboration and pushing the boundaries of scientific inquiry, particularly in biology and biomedical research.
The content above is a summary. For more details, see the source article.