10x Genomics Launches Sentira Single Cell, an Agentic Claude Analysis App

10x Genomics today launched Sentira Single Cell, a cloud-based agentic analysis application for its Chromium single-cell platform, built on Anthropic's Claude models and priced through 10x Cloud Credit packs that run from $1.20 per credit in a 100-credit pack down to $1.00 per credit at the 25,000-credit tier ($25,000).

The company estimates each agentic session consumes roughly 12 credits, and a complete analysis through to experimental conclusions requires around 60 credits. At list prices, that places a full analysis at approximately $60–$72 depending on pack size. All users receive 100 promotional credits at launch for the first Chromium dataset uploaded — enough, the company says, to reach experimental conclusions on one dataset and learn the platform. Volume discounts lower the per-credit cost as pack size increases.

What does Sentira actually do?

Sentira Single Cell runs on 10x Cloud, providing the compute to analyze Chromium data and run multiple analyses in parallel without researchers configuring or managing their own infrastructure. Users interact through a chat interface and ask questions of their data dynamically; on the back end, the system scripts and writes the analysis code.

Michael Schnall-Levin, PhD, CTO of 10x Genomics, frames the tool as the answer to the field's long-standing "last mile problem." Single-cell researchers each bring unique experimental methods and questions, and historically each has required writing custom analysis code. "It has not been scalable to write software that can do everything for everyone," he said. With large language models, researchers can now get answers without being bioinformaticians.

The speed claim is the headline benefit. "What might have taken someone weeks before could take hours," Schnall-Levin said, calling the effect "massively accelerating" for users and, ultimately, for the pace of scientific progress. He also expects secondary benefits: the tool may surface results researchers are not currently seeing in their own data, and 10x hopes it will broaden the pool of people who can query single-cell data directly.

Whose model, and where does the data go?

10x is not training its own LLM. Sentira relies on Claude models, though the platform is not locked to a single model — it can use different LLMs, and the Anthropic models are working well, according to Schnall-Levin. The launch aligns with Anthropic's broader push into the life sciences industry.

On data handling, 10x states it will use researchers' data to understand how the application performs and to improve it, but asserts there is no flow of customer data back into the underlying models.

Sentira is an umbrella brand for 10x software capabilities. Sentira Spatial launched a few weeks ago; today's announcement covers single-cell analysis only.

The project traces to early 2026, building on work 10x did with James Zou, PhD, associate professor of biomedical data science at Stanford University. Zou introduced the "Virtual Lab" — a laboratory made of algorithms — in a 2025 Nature paper, "The Virtual Lab of AI agents designs new SARS-CoV-2 nanobodies."

How reliable is AI-generated analysis?

Schnall-Levin is candid about limits. The tools are accurate, he says, but not perfect. "We don't want someone to just start putting stuff out in a publication with no additional review. We think that this is a really good way to generate initial analysis and exploration, but people are still on the hook and responsible for it," he said. Asked about AI risks broadly, he said (chuckling) that he is "far more nervous about AI risks outside of this."

An internal concern at 10x: as tools grow more powerful, users may rely on them without thinking. Schnall-Levin compares the risk of lost expertise to celestial navigation — few people today can navigate by the stars because technology made it unnecessary. Expertise loss is a double-edged sword, he concedes, but he expects the benefits to outweigh the downside.

A sharper warning comes from outside the company. Luciano Martelotto, PhD, director of global market development for single cell at Waters Biosciences, has criticized AI-driven single-cell analysis in general — not Sentira specifically. "Nothing in an AI-generated analysis looks broken. The code runs, the plots are pretty, the story makes sense. The mistakes that matter are the silent ones," he said. None of those mistakes stem from incompetence, he added: "It does what it's asked. What it doesn't do is get that uneasy feeling when a result looks too good. That feeling comes from having watched data go wrong, and I don't think it can be prompted or coded into existence… not yet, at least."

For R&D managers, the near-term question is budget arithmetic: at roughly 60 credits per full analysis, heavy Chromium shops can model costs directly against the bioinformatics head hours the tool replaces. Schnall-Levin's position is that the direction is set regardless — the single-cell field now has another tool to move with it.