AI has the potential to significantly accelerate the pace of scientific discovery and the development of medical interventions. Since our life sciences efforts began last fall, we have worked to improve the capabilities of our models, build connections to the scientific ecosystem through MCPs and skills, and build partnerships to realize this potential.
Today we’re introducing our most significant extension to these efforts: Claude Science, an artificial intelligence toolkit for scientists. Claude Science is an application that brings together the tools and packages researchers use most, creates verifiable artifacts, and provides flexible access to computing resources.
Introducing Claude Science
Scientific research is often tedious. Researchers must work with dozens of databases, each with their own schema, contend with file formats that require specialized data pipelines and viewers, and switch between a suite of tools: PubMed, Jupyter, R, Cluster Terminal, and others.
Claude Science brings these disparate tools together into a single research environment where scientists can conduct all phases of their work. It helps you analyze literature and perform multi-step research, create detailed artifacts, and allow you to iteratively refine figures and manuscripts until they are ready for publication. Each output contains a verifiable history of how it was created, so you can verify and reproduce the results. As with Jupyter Notebook, you can access Claude Science wherever you’re already working: locally on macOS or Linux, on a remote machine via SSH, or through an HPC login node.
Users interact with a generalist coordinator agent with access to more than 60 curated skills and connectors pre-configured for genomics, single-cell research, proteomics, structural biology, chemoinformatics, and more. These agents can launch others and interact with specialized agents created by users. And the reviewing agent checks quotes and calculations, noting and correcting errors.
Today we are releasing a beta version of Claude Science to Claude Pro, Max, Team and Enterprise users and will continue to improve the platform as we gather user feedback.
How does this work
Claude Science displays proteins, structures and molecules in their natural state, with each result reproduced and traced back to its code.
Rich scientific artifacts, fully reproducible. Scientific research is inherently visual, so Claude Science creates drawings and manuscripts along with the code that created them. It natively visualizes rich scientific artifacts, including 3D protein structures, genome browser tracks, chemical structures, and more. You can talk to your agent about any details, commenting on drawings and manuscripts online so the agent knows where to go to get them ready for publication.
When Claude Science generates a drawing, it includes the exact code and environment in which it was created, a simple description of how it was created, and a complete message history. This allows you to understand the input data, making it easier to check and reproduce the work even months later. You can ask Claude Science to make simple language changes to the numbers—for example, removing grid lines or changing the axis for a logarithmic scale—and the agent will edit its own code.
Claude Science creates environments and manages compute on your laptop, cluster, or GPUs on demand.
Manages your computing and scales on demand. Large analyzes—for example, protein folding or running a genomic pipeline over a huge data set—often require researchers to shift their attention to setting up a computational job, waiting for it to be submitted to a cluster, testing for success or failure, and getting the results back. Claude Science will handle this process for you. It plans, queries before accessing new resources, and allows you to review or reverse any decision before writing and submitting a job to the compute resources your lab is already using (your own HPC cluster over SSH or your modal account for on-demand computing), scaling the analysis from a single GPU to hundreds as needed.
Because its agents run inside a running session that stores the context in memory, even large datasets only need to be loaded once. It runs on your lab’s own infrastructure—your laptop, Linux, or HPC login node—so large or sensitive data sets never have to leave the systems they already reside on, and only the context needed for each analysis step is sent to Claude. As the pipeline runs, the review agent checks the output, noting incorrect quotes, untracked numbers, and numbers that don’t match their base code, and self-corrects along the way. You can fork a session at any time to compare two approaches without losing the original flow.
Claude Science is pre-configured for genomics, single-cell research, proteomics and cheminformatics and is supported by more than 60 scientific databases.
The domain is ready on the first day. Scientific knowledge is scattered across hundreds of specialized sources. For example, in biology, relevant data may be stored in resources such as UniProt, PDB, Ensembl, Reactome, ClinVar, ChEMBL, GEO—each with its own schema and query language—as well as journals and preprint servers, as well as domain-specific open models. When you ask Claude Science a question in layman’s terms, agent specialists query and synthesize all of these sources so you don’t have to navigate them individually. Claude Science leverages the skills of the NVIDIA BioNeMo Agent Toolkit to naturally connect to life science models and libraries in BioNeMo, including Evo 2, Boltz-2, and OpenFold3.
Scientists already have models, datasets, and pipelines they trust. Claude Science can connect to them as well, saving any pipeline as a reusable skill or accessing your lab’s preferred tool through a connector, with future sessions inheriting them automatically. This customization ability allows you to access Claude, your own data, and the trusted tools you already rely on, all in one conversation. Claude Science benefits from our partners’ specialized expertise and platforms, while more scientists access their tools through Claude.
What Scientists Do with Claude Science
Over the past few months, researchers have been working with Claude Science in beta on tasks such as single-cell RNA sequencing analysis, CRISPR screen development, protein structure prediction, chemoinformatics, and more.
Manifold Bio develops tissue-targeted drugs (that target a specific organ or cell type so the drug works where it’s needed and spares the rest of the body) and tests how millions of candidate binders corresponding to hundreds of targets are distributed throughout a living organism at the same time. Manifold used Claude Science to identify targets for its latest experiments. For each tissue and target, Claude Science assessed surface expression, turnover and safety, ranking candidates based on criteria Manifold derived from its own internal data. According to Manifold, what sets Claude Science apart from the average coding assistant is that it can do it holistically, collecting the right data and applying the right judgment, given the built-in context of past programs.
Jérôme Lecoq, a neuroscientist at the Allen Institute, used Claude Science to create a multi-agent “computational review template” that includes about 20 specialized skills for writing long-form reviews. Sub-agents read thousands of documents, extracted the main statement and key quantitative findings and stored them in a government database. The pipeline then creates a narrative arc by recording the review section by section and delegating each section to its specialized subagent. In each section, special agents generate quantitative cross-sectional data directly from the evidence database. A key component of the workflow implemented by Claude Science is the use of actor-critic pairs: one agent creates the content, and the other agent-reviewer evaluates it for accuracy and citation validity.
Before Claude Science, it might have taken Lecoq’s team as long as two years to write such a review. He now has about 10 reviews, many of them over 100 pages, with citations vetted by agent reviewers. The team is currently working with subject matter experts to further improve the AI-based critique agents.
And Stephen Francis, an assistant professor and epidemiologist at the Brain Tumor Center at the University of California, San Francisco, has used Claude Science to support research into the molecular epidemiology of glioma, a type of primary tumor that begins in the glial cells of the brain. His laboratory is investigating the genetic basis of how thousands of germline variants with small effects combine to form individual susceptibility. Although this work predates Claude Science, Francis said the app significantly speeds up the analysis, allowing complex germline studies using different approaches to be performed about ten times faster than before. His team independently verified Claude Science’s results, confirming that it can perform both fast and reliable analysis.
Getting started with Claude Science
Claude Science is available in beta on macOS and Linux for Pro, Max, Team, and Enterprise plans. We’re sharing it early so scientists can start using it to solve real-world problems and tell us how to improve it.
Team and Enterprise users will need an administrator to enable Claude Science. We now have a Team plan, offering discounted space to active research labs in academic institutions and non-profit research organizations; find out more here.
We will also support up to 50 Claude Science AI for Science projects with loans of up to $30,000. Modal will also provide up to $2,000 in computational costs for select projects. We are looking for projects that span multiple fields and explore the frontiers of science, with a primary focus on biology and biomedical research. Applications will be accepted until July 15, 2026, and award notifications will be sent out by July 31. Projects will run from September 1 to December 1, 2026 – apply here.
To stay up to date with product announcements, provide feedback, and learn from others in the Claude Science community, join the AI for Science Discourse.
Get started with Claude Science at claude.com/science.