Your AI assistant forgets your programs. Pluto doesn’t.
Your AI assistant forgets your programs. Pluto doesn’t.
The Pluto MCP server connects your AI assistant to your programs, data, and validated analysis pipelines, so every conversation starts with the context your team has already built.
You open an AI chat to ask a quick question about your program.
First, you paste the broad objective and your hypothesis. Then you explain which of your samples are human and which are cell lines, and you include a few example rows from your metadata to give it the full picture. Ten minutes later, you still haven’t asked the question.
Every new chat starts from zero, and the work is scattered across three places: raw data in one platform, analyses in another, and reports in a third. Unless you copy all of it into the chat, the assistant can’t see your program, and nothing it helps you produce makes its way back to the program record.
With the Pluto MCP connector, the assistant looks that context up itself.
It works with any AI assistant that supports custom MCP connectors, including Claude. Once connected, it can find your organizations, programs, datasets, and analysis spaces. It can run analyses in Pluto when you ask, and it can push external results back into the platform, so they stay with the program instead of sitting in someone’s notebook. When a reviewer asks where a plot came from, the answer is one click away.
What this looks like in practice
Say you’re planning the next phase of a study and can’t remember which samples came next in the series. Without the connector, you’re digging through a lab notebook, searching old records, or cross-checking sample sheets in the platform. With the connector, you ask in plain language and get the answer from your Pluto data.
(//videos.ctfassets.net/a8kqajd3asj2/5gtyLhPUiPAnnn2aY2xAbs/4d5abf4c79a38ba9aacb58f21760a4b7/mcp-next-analysis.mp4)
The video goes a step further. Using a public dataset, we looked at disease signatures in blood, specifically inflammatory arthritis. In two prompts, we went from spotting a gap in the existing analyses, based on a hypothesis, to two new connected plots in Pluto.
“The MCP server works like a scientific co-pilot for my day-to-day work. It shows me what data we have and where the gaps are, and it runs Pluto’s validated analyses across that data from a single prompt. I get the rigor and reproducibility I rely on in Pluto, with fewer steps.”
Caitlin Winkler, PhD, Senior Solutions Scientist, Pluto Bio
The analyses you run through the connector use Pluto’s versioned pipelines with recorded parameters. The same inputs go through the same method every time, and any result the assistant reports shows how it was produced. That matters when a result goes into a program review or a go/no-go discussion.
What your AI provider sees
When your assistant calls a Pluto tool, Pluto returns only the result of that request. That result is then part of the conversation with your AI provider, such as Anthropic or OpenAI, the same way text you type into a chat is. It can stay at the level of program names or go as granular as sample metadata. Retention and training use are governed by your organization’s agreement with the provider.
Your team stays in control
Each user signs in with their own Pluto account, and the assistant sees only what that account can already see. Connector access can be read-only. You can approve each tool call one at a time, and you can disconnect at any time. If you allow write access, the assistant can create new analyses, and each one is logged in your program’s History tab as created by you through your AI assistant, so you always know what was generated and by whom.
Get Started
New to Pluto?
[Book a demo(https://pluto.bio/demo)] and we’ll show you how the MCP server fits your team’s workflow, using your own scientific questions.
Already a Pluto client?
Your admin can add the connector in about two minutes. See our help doc: Connect your AI assistant to Pluto.