← Reference Deep-Dives Reference Deep-Dive · Updated · 9 min read

Run User Interviews from Claude: Connector, Plugin, Prompts

By Kevin, Founder & CEO

Claude can run a real user interview study for you: write the plan, recruit the participants, field AI-moderated interviews, and bring the findings back into the chat with quotes cited to the source interview. This guide covers what is specific to Claude: the connector, the Claude Code plugin and its skills, the cards Claude opens at review, approval and readout, and prompts that put them to work.

How do you add User Intuition to Claude?

In Claude on the web and in Claude Desktop, the connector is all you need: it gives Claude the research tools, and Claude runs the full workflow with it, from brief to cited findings. To add it, open Settings → Connectors in Claude, browse the directory, and search for User Intuition Research. Click Connect and sign in with your User Intuition account, or create a free account first if you don’t have one. Sign-in uses OAuth, so there is nothing to paste and no key to manage. The connector then appears in your Claude chats and in Claude Desktop, so one sign-in covers both. To confirm it works, ask Claude to list your User Intuition studies. From there, describe what you want to learn, and Claude takes it from brief to saved plan, estimate, launch and findings. The Claude directory guide shows each screen.

How do you use User Intuition from Claude Code?

Claude Code works from the terminal and offers two routes. Install the plugin (it includes the hosted server), or add the server alone with claude mcp add --transport http userintuition https://mcp.userintuition.ai/mcp. You need only one of the two.

Option 1: install the research plugin. The user-intuition-research plugin is for Claude Code. It includes the hosted server and packages the research playbook as ten skills:

/plugin marketplace add user-intuition/skills
/plugin install user-intuition-research@user-intuition

Option 2: add the server alone. If you want the research tools without the skills, register only the hosted server:

claude mcp add --transport http userintuition https://mcp.userintuition.ai/mcp

Either way, run /mcp and sign in with your User Intuition account. A Claude Desktop config file that runs the server as a local process uses a USERINTUITION_API_KEY instead; the MCP quickstart has that config.

Each skill covers one research job. You never need to name a skill: ask in plain language and Claude Code picks the right one. Unlike community research skills that stop at a plan, these skills field real interviews with real participants and bring back cited findings. The skills library documents each skill’s steps and inputs.

SkillWhat Claude Code does with itAsk it like this
create-study-from-briefTurns your brief into a saved study plan and relays the planner’s questions”Turn this brief into a study”
design-screenersWrites screener questions that admit the right participants”Screen for people who bought oat milk in the last month”
run-a-concept-testSets up an image concept test or a live-link prototype test with the right probes”Test these two package designs”
field-a-panelGets a panel estimate and launches recruitment after your approval”Field this with 20 US participants”
invite-your-own-participantsSends the study to your own customers by email or link”Invite these 30 shoppers from our loyalty list”
monitor-fieldingReports completes and progress against the target”How is fielding going?“
retrieve-study-resultsPulls the finished report and its sections”Get the report for the pack study”
analyze-completed-studyBuilds findings, themes and evidence from completed interviews”What did we learn, with quotes?“
search-researchAnswers questions across your research library with cited quotes”What have shoppers told us about checkout?“
curate-interview-qualityReviews interview quality inside one study”Which interviews are weakest?”

Concept images from your project. Claude Code can read files on disk, so a concept test can start from designs already sitting in the repository. Name each file and the label participants should see, such as designs/pack-a.png as Pack A, and Claude Code attaches each image to the study under that label.

A three-step example in Claude Code:

  1. Brief. “Turn this brief into a study: test the two front-of-pack designs in designs/ for our granola, labelled Pack A and Pack B, with US shoppers who bought granola in the last month. Recruit from the panel.” Claude Code creates the study, relays the planner’s questions, and shows the saved plan with the image concept test set up.
  2. Field. “Field this with 25 US participants.” Claude Code requests the panel estimate and shows it beside the plan, and recruitment starts once you confirm in the conversation.
  3. Read. “What did we learn, with quotes?” When interviews complete, Claude Code returns findings ranked by prevalence, each with verbatim quotes linked to the source interview.

What does a study run look like in Claude?

Every client runs the same research steps, and the MCP research API reference maps each one. Claude adds cards: because it supports MCP Apps, the saved plan, the approval and the evidence open inside the chat rather than as long blocks of text. The run below is an illustrative shopper study. Where a stage returns findings, it shows example output from a separate User Intuition study of 43 Walmart shoppers, each reconstructing one recent trip in a voice interview, not output the run itself produced.

Step 1: The brief. You write one prompt:

I want to understand how shoppers decide during a single big-box store trip: why they chose the store, where the trip went wrong, and whether anyone fixed it. Recruit from the panel in the US.

Step 2: Planning questions. Claude asks whether to recruit from the panel or invite your own users, creates the study and hands your brief to the User Intuition planner. The planner’s questions, such as how recent the trip should be and whether online orders count, appear in the chat, and Claude passes your answers back unchanged.

Step 3: The review card. Ask to see the study, and Claude opens the saved plan as a card: objectives, screener, and a discussion guide that has each shopper reconstruct one recent trip, then ladders from what they did to why they did it to how it felt. Edit in plain language, such as “add a probe on self-checkout”, and the card shows the new saved version.

Step 4: The approval card. Claude requests a panel estimate for the country you named and shows the plan and the estimate together on one card, with the cost, your balance and the expected timeline. You approve with one click, and recruitment begins. If you change the plan or the audience after that, Claude brings back a fresh estimate to approve, so the version you approved is the version that fields.

Step 5: Fielding. The moderator follows each shopper’s own story and ladders 5–7 levels deep. Ask “how is fielding going?” and Claude reopens the review card with progress against the target. Your team can also watch any interview live in Live view.

Step 6: Findings with citation cards. Ask “what did we learn?” and Claude returns findings ranked by prevalence. Example output from the Walmart study reads like this:

  • 34 of 43 shoppers hit at least one obstacle on the trip.
  • The shelf carried 20 of 51 friction moments across those trips.

Each quote in Claude’s answer carries a citation label. Click it and a card opens inside the answer with the exact quotation, the transcript around it, the study name and, for voice and video, the recording at that moment. One citation from the study, W37: “I preferred figuring it out by myself if I needed help you know, and I couldn’t find it on my own, then I would ask for help.”

Step 7: A follow-up question. Ask something the study raised, such as “what would have made the shelf easier to shop?”, and Claude answers from the interviews with the citation cards attached. When earlier studies cover the same ground, Claude can also ask questions across past research.

Explore the Walmart study for its full findings and sample interview recordings.

Which Claude surface should you run research from?

Claude on the web, Claude Desktop and Claude Code reach the same studies, so the choice comes down to where the question tends to come up.

  • Claude on the web and Claude Desktop suit brand, insights and marketing work. The connector is the whole setup, the cards keep the plan, the approval and the evidence in view, and a finished readout can go straight into a document or an artifact.
  • Claude Code suits product teams who work in a repository. The skills give it a step-by-step research playbook, concept images can come from files in the project, and findings can land in a spec or a pull request description with their citation labels intact.

Studies belong to your User Intuition account rather than to the Claude surface that started them. A study planned in a Claude chat on Monday can be read back from Claude Code later in the week, and both see the same plan, interviews and report.

Which prompts can you paste into Claude today?

  1. Test packaging. “Here are two package designs for our oat milk, labelled Pack A and Pack B. Run a concept test on which one shoppers would pick up first and why, recruit US shoppers from the panel, and open the saved plan for review.”
  2. Test a claim before launch. “Set up a study on whether people believe ‘cut your grocery bill by 20%’ and what proof they would need. Ask me how to recruit, then bring me the plan and estimate to approve.”
  3. Reconstruct a shopper trip. “Interview 20 US shoppers about their last grocery trip: what was on their list, what they swapped at the shelf, and why. Recruit from the panel and show me the saved plan before anything else.”
  4. Understand brand switching. “Plan a study with people who moved to a different coffee brand in the last six months: what triggered the move and what would win them back.”
  5. Draft the readout as an artifact. “Analyze the completed study, rank the findings by prevalence, and build a one-page readout as an artifact that keeps every citation label.”
  6. Test designs from the repository, in Claude Code. “Run a concept test with designs/pack-a.png and designs/pack-b.png, labelled Pack A and Pack B, with US shoppers who bought granola in the last month. Show me the saved plan, then the panel estimate.”
  7. Interview your own users. “Create a voice study on why trial users don’t finish setup in our budgeting app, then help me invite the list I’ll paste.”

One habit makes every one of these sharper: name the decision the study informs. “Lead with taste or with price on the new pack” gives the moderator something to probe toward; “test this design” does not.

Synthetic respondents are also available in chat for a directional first read; the synthetic respondents page explains how they are built.

Where does Claude fit in your research stack?

Claude becomes the interface; User Intuition supplies participants, moderation and evidence. User Intuition’s AI moderator runs 20–30 minute interviews that ladder 5–7 levels deep, with participants from a vetted 4M+ consumer panel (B2B audiences recruited to order per study), or your own participants. Every panelist call is reviewed by the User Intuition team, panelists pass a video identity check, and only interviews that pass quality scoring are billed; see how quality works. The agent shows you the saved plan and a cost estimate before anything is spent; in Claude you approve with one click, elsewhere you confirm in the conversation.

Other agents and products get the same workflow through the research API and MCP server for AI agents, and AI-moderated in-depth interviews explains how the moderator ladders each conversation.

Connect User Intuition in Claude

Add User Intuition Research from Claude’s connector directory, sign in, and paste one of the prompts above. Each voice interview is $30, and signup includes 3 free interviews with your own participants; panel recruiting is billed separately. New to User Intuition? Create a free account first, then come back to Claude and describe what you want to learn.

Once you are connected, pick a first study:

Note from the User Intuition Team

User Intuition provides AI-moderated qualitative research for agencies, consulting firms, and research teams. Keep your methodology and discussion guide, bring your own sample or use our 4M participant panel, and review recordings, transcripts, and evidence-linked findings. Your researchers connect the evidence to the client decision and prepare the final recommendations.

Inspect complete sample calls and a readout, then test your own brief. Starter voice interviews cost $30 with your sample or $60 with standard panel recruitment, with no monthly fee. Specialty audiences are quoted separately; incentives you arrange for your own sample are additional. See pricing or try 3 free voice interviews with your own participants.

Frequently Asked Questions

In Claude on the web or desktop, open Settings, then Connectors, find User Intuition Research in the directory, click Connect and sign in with your User Intuition account; the connector alone runs the full research workflow. In Claude Code, the user-intuition-research plugin brings the hosted server with it, or you can register the server on its own with claude mcp add --transport http userintuition https://mcp.userintuition.ai/mcp.

In Claude Code, the plugin includes the hosted server and ten research skills, including create-study-from-brief, design-screeners, field-a-panel, invite-your-own-participants, monitor-fielding, run-a-concept-test, analyze-completed-study and search-research. The skills field real interviews with real participants, and a plain request such as "turn this brief into a study" runs the right sequence of steps. In Claude on the web or desktop, the connector runs the same workflow on its own.

Claude supports MCP Apps, so when you ask to see a study, its saved plan and fielding progress open as a read-only card inside the chat. Before launch the card shows the exact plan you are about to sign off on; once interviews start, the same card shows progress against the target, so you never leave the conversation to check either one.

Yes. When Claude answers a research question, each quote carries a citation label. Click the label and a card opens inside the answer with the exact quotation, the transcript around it, the study name and, for voice and video interviews, the original recording at that moment.

No. The Claude connector and the hosted server in Claude Code both sign in with OAuth, so there is no key to copy or store. A key from Settings → API Keys, starting with ui_sk_, is only for a Claude Desktop config file that runs the server as a local process.

Start with create-study-from-brief. It turns a plain-language brief into a saved study plan, asks whether to recruit from the panel or your own participants, relays the planner's questions and shows the plan for review. You don't have to name it, because a request like "turn this brief into a study" loads it. The other nine skills take over for screening, fielding, analysis and search as the study moves on.
Get Started

Put This Research Into Action

Run your first 3 AI-moderated customer interviews free with your own participants — no sales call.

Self-serve

Launch your first study in minutes. Results in 24 hours.

See it First

Explore a real study output — no sales call needed.

No contract · No retainers · Panel results in 24 hours