Anthropic has taken its biggest step yet into the life sciences with Claude Science, an AI workbench for researchers announced June 30, 2026. The app folds the tools scientists already use — PubMed, Jupyter, R, a cluster terminal — into a single environment where a coordinating agent runs literature analysis, multi-step research, and publication-ready figure generation with full reproducibility.
What Claude Science Does
Claude Science starts from a generalist coordinating agent with access to more than 60 curated skills and connectors pre-configured for genomics, single-cell, proteomics, structural biology, and cheminformatics. It can spin up specialist agents, and a reviewer agent checks citations and calculations, flagging and correcting errors as work proceeds.
Every output carries an auditable history: when the agent generates a figure, it includes the exact code and environment that produced it, a plain-language description of how it was made, and the full message history — so results can be validated and reproduced months later.
NVIDIA BioNeMo Built In
Researchers can ask questions in plain language and specialist agents synthesize across hundreds of sources rather than forcing them to navigate each schema, including UniProt, PDB, Ensembl, Reactome, ClinVar, ChEMBL, and GEO. The platform uses the NVIDIA BioNeMo Agent Toolkit to connect natively to life sciences models and libraries, including Evo 2, Boltz-2, and OpenFold3.
Compute Where Your Data Lives
Because the agents work inside a running session that holds context in memory, massive datasets are loaded only once, and analysis runs on the lab’s own infrastructure — a laptop, Linux box, or HPC login node — so large or sensitive datasets never leave the systems they are already on. For heavier jobs, on-demand GPU capacity via a Modal account lets a single analysis scale from one GPU to hundreds.
Real-World Results
- Manifold Bio used Claude Science to nominate targets for its tissue-targeting medicine experiments, scoring surface expression, trafficking, and safety end to end
- The Allen Institute’s Jérôme Lecoq built a multi-agent computational review pipeline that turned multi-year review writing into roughly 10 reviews of 100-plus pages each
- UCSF epidemiologist Stephen Francis reported germline variant analyses in about one-tenth the previous time, with results independently validated
Availability
Claude Science is available in beta on macOS and Linux for Pro, Max, Team, and Enterprise plans, with Team admins enabling access. Anthropic will support up to 50 AI for Science projects with up to $30,000 in credits, and Modal is providing up to $2,000 in on-demand compute for select projects. Applications close July 15, 2026.