Anthropic Launches Claude Science and Drug Program

Anthropic announced Claude Science on June 30, 2026, describing it in its own blog post as an AI workbench that integrates the databases, notebooks, and compute environments researchers already use, including PubMed, Jupyter, R, and HPC clusters, into a single auditable workspace. Every output carries a traceable history of the code and data that produced it, and a generalist coordinating agent has access to more than 60 curated skills and connectors pre-configured for genomics, single-cell analysis, proteomics, structural biology, and cheminformatics, built in part on NVIDIA's BioNeMo toolkit. The app is available in beta on macOS and Linux for Claude Pro, Max, Team, and Enterprise users. At a San Francisco launch event covered by MIT Technology Review and STAT, Anthropic's head of life sciences, Eric Kauderer-Abrams, and other executives said the company will also begin its own drug-discovery programs, focused on neglected diseases that Kauderer-Abrams said traditional pharma and biotech consider commercially unattractive despite real disease burden.

MIT Technology Review reports the launch follows Anthropic's April acquisition of Coefficient Bio, an eight-person computational-biology startup founded by former Genentech scientists, for roughly $400 million in stock, and that Claude Science was demonstrated identifying drug candidates for phenylketonuria, a rare genetic disease, at the launch event. STAT and MIT Technology Review both note executives were candid that it remains unclear whether Anthropic intends to commercialize any resulting drug candidates; the stated goal is firsthand validation experience rather than a near-term pipeline. For teams evaluating the product, the auditable-history feature and pre-built domain connectors target two real adoption blockers in regulated science, provenance of AI-assisted results and integration friction with existing lab tools, though neither substitutes for assay-level or clinical validation.

The launch positions Anthropic alongside Google DeepMind, whose AlphaFold work won a Nobel Prize, as a leading AI-for-science player; MIT Technology Review notes DeepMind researcher John Jumper announced this month that he is moving to Anthropic. Anthropic frames Claude Science as its third flagship product alongside Claude Code and Claude Cowork, and MIT Technology Review notes that deepening pharma relationships could help the company toward the profitability and IPO timeline it has previously signaled.

Watch for whether Anthropic publishes reproducible benchmarks tying Claude Science outputs to experimental validation, for results from named early users the company cites, including Manifold Bio, an Allen Institute researcher, and a UCSF Brain Tumor Center researcher, and for any disclosure on whether the internal neglected-disease programs are intended to reach clinical development or remain a validation exercise.

This fits a broader industry pattern of AI vendors trying to prove domain value by doing the work themselves rather than only selling tools, similar to how coding-assistant vendors dogfood their own products; it is not a specific claim about Anthropic's commercial intentions beyond what its executives have stated.