Anthropic introduced Claude Science, a beta research workbench for scientists, on June 30, giving Claude Pro, Max, Team, and Enterprise users access to an integrated application for literature review, data analysis, figure generation, and computational workflows. The release comes one week after NVIDIA unveiled its BioNeMo Agent Toolkit, marking a push to move AI assistants beyond general productivity chat and into auditable, reproducible scientific research.
The launch reflects ongoing investment in AI for Science & Research, where the bar for tools is rising from generating plausible answers to producing results that can be traced, challenged, and repeated within established review processes. Anthropic said Claude Science is available on macOS and Linux, and can run locally, on remote machines over SSH, or through a high-performance computing login node.
Inside the workbench
Claude Science pulls fragmented research tools into a single environment. It connects to PubMed, Jupyter, R, cluster terminals, and domain-specific databases, while preserving a complete auditable history of how outputs were produced. A generalist coordinating agent manages more than 60 curated skills and connectors configured for genomics, single-cell analysis, proteomics, structural biology, cheminformatics, and other disciplines. Users can also build specialist agents, and a built-in reviewer agent checks citations and calculations, flagging errors.
The workbench emphasizes reproducibility. When Claude Science generates a figure, it includes the code and environment used, a plain-language description of the process, and the message history leading to the output. Anthropic said the preserved history is intended to make results easier to validate and reproduce later, answering a key demand from research teams that need to audit AI-assisted work.
NVIDIA's BioNeMo connection
NVIDIA's BioNeMo Agent Toolkit, announced June 23, supplies domain-specific tools and skills for agentic life sciences workflows. It includes NVIDIA life sciences libraries, open models like Evo 2, Boltz-2, and OpenFold3, and is designed to help agents gather evidence, reason across findings, run computational experiments, and recommend next steps. Claude Science uses BioNeMo Agent Toolkit skills to connect to these life sciences models and libraries, deepening its access to specialized research infrastructure.
Why this matters for scientists and researchers
For research groups and enterprises evaluating AI tools, the key test will be whether agentic AI can produce reliable, traceable results while keeping humans in control of sensitive data, compute decisions, and scientific judgment. Claude Science and the BioNeMo Agent Toolkit aim to address that by embedding audit trails, preserving computational environments, and connecting to established lab tools directly. The approach shifts the AI from a black-box assistant to a piece of the documented scientific record.
Your membership also unlocks:
