Sprig MCP connects your survey data to the AI workspace you already live in. A month after launch, four use cases keep showing up: planning studies in Claude before they touch Sprig, running crosstabs in plain language, doing advanced pricing and feature analysis without a stats package, and pulling themes into the tools where decisions actually get made.
What is Sprig MCP?
Sprig MCP is a Model Context Protocol connector (mcp.sprig.com/mcp) that links Sprig survey data to AI assistants including Claude, ChatGPT, Gemini, and Copilot. It has two modes: Create, where you plan studies in your LLM and build them in Sprig, and Analyze, where you pull surveys, responses, and themes into your LLM. It launched in June 2026 and is admin-enabled, with your existing role-based access and PII controls carried over.
When we launched this feature, I expected the Analyze tooling to dominate. Instead, teams have jumped on both halves of the connector at once. Here’s what a month of real usage looks like.
1. How do teams plan studies with Claude before building in Sprig?
Researchers have always sketched studies outside of their research tool first, whether on a document, a whiteboard, or a Slack thread with a colleague. What’s changed is that the sketching conversation now happens with AI that can hand the finished plan directly to Sprig.
One pattern I keep seeing: a PM describes the decision they need to make (“we’re choosing between two onboarding flows”), works with Claude on what to ask and who to ask, pressure-tests the questions for bias, and then sends the plan into Sprig where the Design Agent programs the full study. The thinking happens in the chat, and the rigor happens on the platform, so nobody rebuilds anything by hand.
2. Can you run crosstabs on survey data in plain language?
Your survey says satisfaction is 7.8. Fine. Is that true for enterprise customers on the new plan? For users who joined in the last 90 days? A crosstab answers the question behind the question, which used to mean exporting to a spreadsheet and building pivot tables.
Now the request is a sentence: “Pull the responses from our Q2 satisfaction survey and compare scores across plan tier and account age. Flag any cell with too few responses to trust.” The Responses endpoint returns up to 1,000 responses per call, and the analysis happens right in the conversation. We published a full prompt guide for this, including how to keep yourself honest about thin cells and multiple comparisons.
3. How do you run MaxDiff, conjoint, or Van Westendorp without a stats package?
These analysis methods earn their reputation for rigor, and also their reputation for requiring either specialized software or a patient analyst.
With survey data flowing through MCP, the analysis step collapses. Field the study in Sprig, then ask your LLM to run the utilities, build the price sensitivity curves, or simulate preference share. You still need to know what the method is for and what its limits are. That knowledge is the researcher’s job and always has been. But the barrier between “we should run a MaxDiff” and “here are the results” is now measured in minutes.
4. How do survey insights get into roadmap conversations?
The quietest use case might be the most important one: teams are pulling survey themes into the workspaces where their roadmap conversations already live. Themes from an open-ended feedback question show up in the same chat where the team is debating what to build next, with traceable quotes behind every claim.
This is the reason we built MCP in the first place. Research that stays in the research tool gets read by researchers, but research that travels to where decisions get made actually changes them.
The part your security team will ask about
MCP access is admin-only, and your existing role-based access and PII controls carry over: if a user can’t see a field in Sprig, they can’t pull it through MCP either. Connecting an LLM to your research data doesn’t mean opening the vault. It means extending the permissions you already set.
Where this is heading
As AI moves researchers from being in the loop to being on the loop, the teams getting the most out of MCP are the ones who already knew what questions mattered and just removed everything standing in between the question and the answer. Allow work to scale as decisions are made faster, without losing their research rigor.
Frequently asked questions
Which AI tools work with Sprig MCP? Claude, ChatGPT, Gemini, Microsoft Copilot, and Cursor. Any MCP-compatible client can connect at mcp.sprig.com/mcp.
Is Sprig MCP secure for enterprise data? Yes. Access is admin-enabled only, and Sprig’s role-based access controls and PII settings apply to every MCP call. Sprig is SOC 2 Type II, HIPAA, and GDPR compliant.
What data can I pull through Sprig MCP? Three endpoints: Surveys (study metadata and structure), Responses (up to 1,000 per call), and Themes (AI-synthesized findings with source quotes).
Connect your workspace at mcp.sprig.com/mcp, or start with our prompt guides for crosstabs, MaxDiff, conjoint, and Gabor-Granger.
