The User Intuition MCP server brings qualitative research into the AI conversation: design a study, approve recruitment, and retrieve interviews and cited findings.
This launch article’s setup and capability reference reflect the current published package, @userintuition-ai/mcp.
What is available now?
The package supports hosted Streamable HTTP with OAuth and local stdio with an API key. It also includes a CLI for the same research tools, browser login, and API-key management. Research skills guide agents through planning, fielding, monitoring, and analysis.
What can an agent do with it?
- Plan: draft a study from a brief, work through the platform’s planning questions, and show the saved discussion guide.
- Price and launch: quote the panel cost and timeline for your audience and country, then field the study after your one-click approval.
- Invite your own people: send study invitations to your customers by email or shareable link.
- Watch progress: follow completed interviews, open transcripts and recordings, and receive webhooks as interviews finish.
- Report: pull findings with prevalence and verbatim quotes, and turn a study into a presentation.
- Check what panels can reach: look up country and language combinations and request feasibility for harder audiences.
Study types are in-depth interview, concept test (image), and prototype test (URL), in voice, chat, or video, in 80+ languages. Participants are consumers from User Intuition’s vetted 4M+ panel or your own customers, and synthetic respondents are also available for a directional first read in chat.
Ask the agent what earlier studies found before briefing a new one: it answers from your research history and cites the interview behind each point.
How do you connect your agent?
Claude users add User Intuition Research from the connector directory, and the MCP quickstart has the command or URL for Claude Code, Codex, ChatGPT, Cursor, and local stdio clients.
Why does agent-native research matter?
A product agent can ask for evidence before writing a feature specification. A marketing agent can test a claim before drafting a campaign. A strategy agent can commission interviews when public sources leave a decision-critical question unanswered.
The researcher stays in the conversation to choose the audience, review the saved plan, and approve recruitment. The platform handles interview delivery and analysis. This makes research a practical step in an agent workflow while retaining the decisions that need human judgment.
Your first study
Tell the agent the decision you face and who you want to hear from, and pick the panel or your own participants. The agent drafts the study, brings you the platform’s planning questions, and shows you the saved discussion guide. For a panel study it adds the price and timeline from estimate_panel, and one click in Claude or ChatGPT launches it.
Results come back as findings with prevalence (n of N) and verbatim quotes linked to the source interview and, for voice and video studies, to the moment in the recording. get_study_report fetches the report, and get_interview opens any interview behind it. For example output, explore the Walmart study: a User Intuition study of 43 Walmart shoppers in which 36 of 43 will shop there again and the headline finding reads “This is not loyalty. It is tolerance, and it is priced.”
Every study the agent runs also appears in your dashboard, where the Customer Intelligence Hub lets your team explore the full research library.
Pricing
Connecting the MCP server is free, with no subscription. Studies use platform credits: $15 per chat interview, $30 per voice interview, and $60 per video interview. Panel recruiting brings a voice interview to $60 all-in, and the agent shows that total in the estimate before you approve. New accounts get 3 free interviews.
See current pricing and the MCP server and research API for AI agents.