Anthropic on June 30 launched Claude Science, an AI workbench designed to help researchers in biology, chemistry, and related fields run experiments, analyze data, and prepare manuscripts inside a single environment.
What it does
Unlike general-purpose Claude, Claude Science integrates with more than 60 curated scientific databases and pre-configured toolkits covering genomics, proteomics, structural biology, cheminformatics, CRISPR screen design, and molecular epidemiology. The platform natively renders 3D protein structures, genome browser tracks, and chemical structure diagrams alongside the reproducible code that generated them.
A coordinating AI agent manages the workflow, creating specialist sub-agents and delegating tasks, while a dedicated reviewer checks citations and validates calculations before results are finalized.
Compute access
Claude Science can connect to a researcher’s existing high-performance computing cluster via SSH or to Modal for on-demand GPU jobs. Sensitive datasets remain on local systems and are not sent to Anthropic’s servers unless the researcher initiates the transfer. The platform also integrates NVIDIA’s BioNeMo toolkit, providing access to biology-focused models including Evo 2, Boltz-2, and OpenFold3.
Who it is for
The tool is available in beta to Claude Pro, Max, Team, and Enterprise subscribers on macOS and Linux. Anthropic offers a discounted Team plan for academic institutions and nonprofits.
Separately, the company is funding up to 50 research projects through a grants program, providing up to $30,000 in credits plus $2,000 in Modal compute per project. Applications are open through July 15, with award notifications by July 31; projects would run from September through December 2026.
Early results
Several research teams piloted the tool ahead of launch. Manifold Bio used it to evaluate tissue-targeting drug candidates. Neurobiologist Jérôme Lecoq said it compressed a two-year literature review into weeks. A UCSF epidemiologist reported cutting germline analysis time to one-tenth of what it previously took.
Anthropic CEO Dario Amodei described the product’s ambition as helping scientists “make sense of that complexity, in its full complexity, better,” while acknowledging results are not guaranteed: “We don’t know for sure if that’s going to work out. But I think we’re seeing signs that we’re seeing the beginnings of it.”
Competitive context
Claude Science enters a field that already includes OpenAI’s GPT-Rosalind, a specialized research model available under enterprise agreements, and Google DeepMind’s Gemini for Science platform alongside proprietary biology models such as AlphaFold. Anthropic frames its approach around workflow integration rather than a new underlying model: Claude Science runs on existing Claude models, including Claude Opus 4.8.