Bring Claude Science into your lab
We set up Claude Science on your research data and computing, and train the scientists who’ll use it.
What Claude Science changes
Most of a scientist’s week isn’t spent on science. It goes to moving data between databases that don’t talk to each other, writing one-off scripts to open unusual file formats, submitting cluster jobs, and checking whether they ran.
Claude Science, Anthropic’s research app, takes on that work. Your researchers can search the literature, run analyses, and draft findings in one place. Every figure comes with the code behind it, a plain-language note on how it was made, and the full conversation that led there. A background reviewer flags any claim it can’t trace to evidence before the result reaches your team.
What it can’t reach on its own is your data in Benchling, your internal databases, and the pipelines your team spent years validating. We connect those, set up a review process your scientists trust, and work with them to find where Claude Science helps their research.
Human-in-the-Loop
Scientists are trained to ask why a result came out the way it did, and our training starts there. We teach your team to supervise Claude Science the way they’d supervise a new postdoc. They review the code, rerun the steps, and test another approach whenever something looks off.
Anthropic’s own researchers work the same way. When they used Claude Science to discover a new enzyme system in September, Claude screened more than 200,000 reverse transcriptases and proposed candidates. Their scientists vetted every one and did all of the lab work.
Claude Science also has some checks of its own. It asks for approval before it uses new computing resources, and you can undo any decision it makes. We add sign-off points on top of that, so a scientist approves the work before it goes into a report or a regulatory submission.
Your data stays where it is
You install Claude Science where your data already lives, whether that’s a lab Linux server, a high-performance computing (HPC) login node, or a cloud virtual machine. Your raw data and compute stay on your own systems. Only what’s in the conversation itself goes to Anthropic, and it’s handled under their standard retention terms.
On the Enterprise plan, your admins decide who gets access, which connectors they can use, and what computing is available to each group. We set that up to match your security and compliance policies, so IT and quality teams can see how it’s being used.
How we work with your team
We usually work in three stages.
Train your scientists
We run hands-on sessions built around your team’s active research questions. Scientists learn how to direct the agent, evaluate its results, and tell when one needs a second look. We start with one workflow your team already runs.
Connect your data
Claude Science already connects to more than 60 public scientific databases. We build connectors to your own lab databases and electronic notebooks. Then we set it up to run jobs on your Slurm cluster or cloud computing.
Make it repeatable
Once your team validates an analysis, we save it as a reusable skill. The next scientist who runs it starts from the version your team already checked. Your research groups end up running the same analysis the same way.
What teams are using it for
- 01
Single-cell and genomics analysis
Clustering and annotating cells across tissues and finding marker genes, with every figure traced back to the code that made it.
- 02
Target nomination
Pulling expression and safety data for each candidate, then ranking them against the criteria your team has learned from past programs.
- 03
Protein structure
Pulling predicted structures, layering on domains and clinical variants, and exploring the model in 3D.
- 04
Cheminformatics and molecular design
Searching bioactivity data, computing properties and similarity, and drawing or refining structures in a live 2D sketcher.
- 05
Literature review
Reading thousands of papers, pulling the key finding from each, and drafting a review whose citations a reviewer agent has checked.
- 06
Program and indication dossiers
Building a fully sourced case for a program, so leadership can see the evidence behind the recommendation.
Proofs of concept with biotech teams
Claude Science is new, and we are currently running proofs of concept with select biotech companies across areas such as computational biology, target discovery, and literature synthesis.
Each engagement begins with a single workflow your scientists run today. We set up Claude Science on your data, train your team members on responsible workflows, and review the outputs alongside them over several weeks. By the end of the program, your team knows exactly how the platform performs on your research, giving you a foundation to build on.
Frequently Asked Questions
How do I know Claude didn’t make a result up?
Every figure and table comes with the code that produced it, the environment it ran in, and a record of how it was made. A background reviewer flags citations it can’t verify and numbers it can’t trace. Our training shows scientists how to check that record the same way they’d check a colleague’s work.
Can we reproduce an analysis without Claude?
Yes. The code and environment are saved with each result, so anyone on your team can rerun it themselves.
We already have validated pipelines. Do we have to rebuild them?
No. We connect Claude Science to the tools and pipelines you already trust, and save them as skills it can reuse.
What do we need to get started?
Claude Science is in beta for macOS, Windows, and Linux. For our work, your team needs to be on a Claude Team or Enterprise plan, since that’s where an admin can switch it on for a group and manage who uses it. If your IT and quality teams need custom roles, usage reporting, and the Compliance API, that’s the Enterprise plan. We can help you work out which one fits. Most teams then start with a small proof of concept on one workflow.
Will Claude refuse some of our drug discovery questions?
Sometimes, yes. Claude’s default safeguards block some biology and drug development requests. In September, Anthropic opened its Life Sciences Verification Program, where labs, biotechs, and pharma companies can apply to have those limits adjusted for their work.
How is this different from your other Claude implementation work?
Our broader implementation work connects Claude to business systems like CRMs and ERPs. Claude Science work is for research teams, and deals with lab data, computing clusters, and scientific analysis.
Bring Claude Science into your R&D pipeline
From connector setup to scientist onboarding, we make sure your AI rollout is secure, repeatable, and tailored to your existing lab infrastructure.