Anthropic ships Claude Science, a research workbench wired into 60 scientific databases

Anthropic released Claude Science in beta this week, and the framing matters more than the feature list. It's an AI workbench that connects to more than 60 scientific databases and acts as a project manager across them, pulling up 3D protein structures, genome browser tracks, and chemistry drawings, then generating figures alongside the code that produced them. It runs on existing models, including Opus 4.8, with no special access.

Anthropic went out of its way to say what this isn't. Not a new model. Not a more capable model for biology. That's a real distinction, and one most launches blur on purpose. I build integrations for a living, and the unglamorous truth is that most of the time a domain doesn't need a smarter model, it needs the model wired into the right tools with the right context. A protein viewer and a genome browser aren't reasoning problems, they're plumbing problems. Claude Science looks like Anthropic betting the plumbing was the bottleneck, not the intelligence.

The credits are the tell

Selected biology and biomedical projects can get up to $30,000 in Claude credits plus $2,000 in Modal compute, with applications closing July 15. That's Anthropic subsidizing adoption in a field where budgets are thin and switching costs are real. Researchers don't rebuild their workflow around a chat window, they rebuild it around a tool that already speaks their file formats.

If you're building anything that touches scientific or technical data, the lesson isn't 'add AI.' It's connect the model to the actual tools your users already trust, and be honest with them about what changed and what didn't. Overclaiming capability is how you lose a technical audience in one sentence.