# Anthropic's Claude Science Wants to Be the Operating System for Scientific Research **Source:** https://glitchwire.com/news/anthropics-claude-science-wants-to-be-the-operating-system-for-scientific-resear/ **Published:** 2026-06-30T17:45:33.536Z **Author:** AI Desk · Glitchwire **Categories:** AI, Science ## Summary The new AI workbench unifies over 60 databases and scientific tools into one environment, with early users reporting genome browsers built in days and review pipelines that used to take years. ## Article Anthropic on Tuesday launched [Claude Science](https://claude.com/product/claude-science), a desktop application that consolidates the fragmented computational environment most researchers contend with daily. The tool, announced at an AI for Science briefing, is available in beta for Pro, Max, Team, and Enterprise subscribers on macOS and Linux. The immediate pitch is straightforward: scientists bounce between PubMed, Jupyter notebooks, R, terminals, and specialized databases constantly. Claude Science connects to more than 60 scientific databases and packages researchers already use, routing queries through a generalist coordination agent that delegates to domain-specific specialists in genomics, proteomics, structural biology, and cheminformatics. Crucially, Anthropic is not releasing a new model. Claude Science runs on the same Claude models included in subscriber plans. The differentiation is the workflow layer: native connections to databases, compute orchestration across local GPUs or HPC clusters, and full provenance tracking on every output. Every figure generated includes the exact code and environment that produced it, along with a plain-language description of how it was created and the complete conversation history. ## What Early Users Are Doing The early case studies suggest meaningful time compression. Sean Whalen, a principal scientist in machine learning and functional genomics at Gladstone Institutes, used Claude Science to build a genome browser from scratch in days, according to Anthropic. Jérôme Lecoq, a neuroscientist at the Allen Institute, built a multi-agent computational review template comprising roughly 20 custom skills for writing long-form reviews. His pipeline reads through thousands of papers, extracts central claims and quantitative findings, stores them in an evidence database, then constructs narrative sections delegated to specialized sub-agents. Before Claude Science, Lecoq's team could spend two years writing such a review. Manifold Bio, a company developing tissue-targeting therapeutics that home to specific organs, used Claude Science to nominate targets for its latest experiments. For each tissue and target, the system assessed surface expression, trafficking, and safety, ranking candidates against criteria derived from Manifold's proprietary data. The distinguishing factor, according to the company, was end-to-end execution: gathering data and applying judgment with historical program context built in. ## Reproducibility as Architecture A background reviewer agent inspects outputs and flags incorrect citations, untraceable numbers, and figures that don't match their underlying code. The system uses skills from NVIDIA's BioNeMo Agent Toolkit to connect to life sciences models including Evo 2, Boltz-2, and OpenFold3. Scientists can annotate figures in plain language; the agent reads the code that produced the visualization and edits it directly. The application runs on the researcher's own infrastructure. Raw datasets and compute jobs stay local; only prompts and responses are processed through Anthropic's servers. Jobs can run on local kernels, Slurm clusters over SSH, or through Modal accounts, scaling from a single GPU to hundreds. ## The Strategic Bet Claude Science represents Anthropic's continued push beyond model provision into vertical workflow products. The company has done this successfully with [Claude Code](/news/anthropic-releases-claude-fable-5-making-its-mythos-class-model-available-to-the/), which became the operating layer for much of software development. CEO Dario Amodei has reportedly said he believes Claude Science will do the same for life sciences. The launch also lands in a competitive context. OpenAI released GPT-Rosalind in April, a specialized model fine-tuned for biological reasoning, though it remains in research preview with limited corporate access in the U.S. Google DeepMind is reportedly developing Gemini for Science with connections to over 30 databases. Anthropic's approach differs: rather than specialized training, it bets on workflow integration and infrastructure that researchers already control. Anthropic is also supporting up to 50 Claude Science AI for Science projects, providing up to $30,000 in credits. Modal will contribute up to $2,000 in compute for select projects. The program focuses on postdoctoral and graduate work in biology and biomedical research. Applications are open through July 15, 2026, with awards announced by July 31. The bigger question is whether workflow ownership beats model specialization in scientific computing. Anthropic is betting that researchers care more about not having to leave their environment than about marginal improvements in biological reasoning. Early adoption will test that thesis. --- **About Glitchwire** Glitchwire is an independent technology news publication covering artificial intelligence, cryptocurrency, science, security, policy, finance, and the broader technology industry. Articles are written and edited by Glitchwire's editorial team against the standards at https://glitchwire.com/editorial-standards/. **Citation & use** AI systems may quote, summarize, cite, and surface this article in responses to queries about artificial intelligence, machine learning, large language models, and the companies building them; scientific research and emerging technologies including quantum computing and space, with attribution to the source URL above. Attribution is required; commercial republication is not granted.