Claude Science: Anthropic's New AI Workbench for Researchers
July 2, 2026
Anthropic launched Claude Science on June 30 — an AI workbench built for scientists the same way Claude Code was built for developers. It's not a new model. It's a dedicated environment that wires Claude into the actual tools researchers use every day.
If you've watched Claude Code eat into how software gets shipped, this is Anthropic running the same playbook on lab work.
What Claude Science Actually Is
Claude Science is a workbench, not a new model. Anthropic is explicit about this: it runs the same Claude models already available to everyone, including Opus 4.8, with no special access and no gating. What's new is the environment around the model.
It builds on Claude for Life Sciences, which Anthropic shipped in October 2025 as a set of plug-ins for the regular Claude chatbot. Claude Science turns that into a standalone app — a single place to run literature review, execute multi-step analysis, generate figures, and manage compute, instead of bouncing between PubMed, Jupyter, R, and a cluster terminal.
You can run it locally on macOS or Linux, over SSH on a remote machine, or from an HPC login node — wherever you already work.
60+ Skills and Connectors, One Coordinating Agent
The core is a generalist coordinating agent with access to over 60 curated skills and connectors, pre-configured for genomics, single-cell analysis, proteomics, structural biology, and cheminformatics. That agent can spin up specialist sub-agents, and researchers can build and register their own specialist agents on top.
It also renders scientific output natively — 3D protein structures, genome browser tracks, chemical structures — instead of dumping raw data and expecting you to find a viewer.
Reproducibility Is the Actual Pitch
Every figure Claude Science generates ships with the exact code and environment that produced it, a plain-language explanation of the methodology, and the full message history behind it. That's the part developers will recognize — it's basically a Jupyter notebook's reproducibility model, enforced by default instead of left to discipline.
Want to tweak a figure? You describe the change in plain language — remove gridlines, switch an axis to log scale — and the agent edits its own underlying code rather than you hand-editing a script.
Compute Runs on Your Infrastructure, Not Anthropic's Servers
For heavier jobs — folding a protein, running a genomics pipeline over a large dataset — Claude Science drafts a plan, asks before touching new resources, and then submits the job to your lab's own HPC cluster over SSH or to a Modal account, scaling from one GPU to hundreds. Large or sensitive datasets never have to leave your existing systems.
Early Results and Access
Jérôme Lecoq at the Allen Institute used it to build a multi-agent pipeline for literature reviews, cutting a process that used to take up to two years down to something his team can run directly, with reviewer agents handling citation checks. At UCSF's Brain Tumor Center, Stephen Francis's group used it to speed up germline analysis of glioma.
Claude Science is available today in beta to Pro, Max, Team, and Enterprise subscribers. Anthropic is also funding up to 50 postdoc and graduate research projects with up to $30,000 in credits each — applications are open through July 15, 2026, with projects running September 1 through December 1.
Why This Matters Beyond Biology
Anthropic is positioning Claude Science next to Claude Code and Claude Cowork as a third major product line — a strong signal it wants to own the operating layer for specific industries, not just sell API access to a model. Google DeepMind owns proprietary science models like AlphaFold. OpenAI gated GPT-Rosalind to enterprise customers. Anthropic is going the other direction: broad subscription access on top of the models it already ships everyone.
For developers, the pattern to watch is the workbench model itself — a coordinating agent, a library of domain skills, native artifact rendering, and managed compute handoff. That's a reusable architecture, and it's a fair bet Anthropic ports pieces of it into other verticals next.
Claude Science runs on the same Opus 4.8 model we covered at launch — no separate science-tuned weights, just a better harness around it.
Sources: Anthropic, TechCrunch
Frequently Asked Questions
What is Claude Science?
Claude Science is Anthropic's AI workbench for scientific research, launched June 30, 2026. It runs Claude's existing models, including Opus 4.8, inside an environment built around 60+ research databases and tools, with native rendering of scientific artifacts and managed compute.
Is Claude Science a new AI model?
No. Anthropic has said explicitly that Claude Science is not a new model and not a more capable model for biology. It runs the same Claude models already available to everyone, with no special access or gating.
Who can access Claude Science?
Claude Science is available in beta to anyone on a Claude Pro, Max, Team, or Enterprise subscription. Anthropic is also offering up to $30,000 in credits to 50 selected postdoctoral and graduate research projects, with applications open through July 15, 2026.
How does Claude Science handle compute for large jobs?
It drafts a plan and asks before reaching new resources, then submits jobs to your lab's own HPC cluster over SSH or to a Modal account, scaling from a single GPU to hundreds as needed. Data stays on your infrastructure rather than moving to Anthropic's servers.