Anthropic launches Claude Science for drug discovery

Anthropic announced Claude Science on June 30, 2026 and says the app is now in beta for Claude Pro, Max, Team, and Enterprise users. The company's announcement describes it as an AI workbench for scientists that can analyze literature, run multi-step research workflows, generate figures and manuscripts, and preserve an auditable history for each result. The app runs where researchers already work, including macOS, Linux, remote machines, SSH, and HPC login nodes.

The official launch says Claude Science uses the same Claude models rather than a specialized biology model. Its differentiation is the surrounding system: a coordinating agent, specialist agents, a reviewer agent for citations and calculations, access to over 60 curated skills and connectors, and native support for scientific artifacts such as protein structures, genome browser tracks, and chemical structures. Anthropic says the app can connect to resources such as UniProt, PDB, Ensembl, Reactome, ClinVar, ChEMBL, GEO, NVIDIA BioNeMo tooling, and lab-specific pipelines.

For computational biology, ML platform, and data-science teams, the core evaluation question is whether Claude Science can make analysis reproducible enough for team review. The important claims are provenance, local or lab-controlled compute, reusable skills, and reviewer-agent checks, not autonomous drug discovery. Teams should test whether outputs can be rebuilt from the stored code and environment, whether source citations survive adversarial review, and whether sensitive datasets remain inside approved compute boundaries.

TechCrunch frames the product as Anthropic trying to own more of the scientific workflow layer, while The Verge reports that Anthropic also plans to pursue its own drug programs for neglected diseases. Blogspan and The Decoder report that Anthropic cited a demonstration in which Claude analyzed 100 rare genetic diseases in under an hour and flagged 32 candidates for computational screening. Those are screening claims, not clinical evidence, so the safer read is that Claude Science may accelerate early triage while wet-lab validation, toxicity work, and clinical trials remain the bottleneck.

The next credibility test is independent validation outside launch-stage examples. Watch for published benchmarks from labs using Claude Science on reproducible workflows, public details on Anthropic's neglected-disease programs, and evidence that reviewer agents reduce citation or calculation errors in real research settings. For regulated or patient-linked data, the strongest adoption signal would be documented governance around retention, audit trails, and deployment on controlled infrastructure.