Claude MCP Customer Research: Setup Guide

Claude can coordinate customer research through User Intuition’s MCP server: shape a study, obtain plan and recruitment approval, and bring interview evidence back into the conversation.

Connect your agent

Use the hosted endpoint, https://mcp.userintuition.ai/mcp, with OAuth in compatible clients. For local stdio, run npx -y @userintuition-ai/mcp with USERINTUITION_API_KEY set to your ui_sk_ key. The CLI supports browser login and API-key management. Hosted OAuth does not require exchanging a user-supplied API key.

For a local stdio client that uses an MCP configuration file:

{
  "mcpServers": {
    "userintuition": {
      "command": "npx",
      "args": ["-y", "@userintuition-ai/mcp"],
      "env": {"USERINTUITION_API_KEY": "ui_sk_your_key_here"}
    }
  }
}

Keep real keys out of version control. For shell workflows:

npm install -g @userintuition-ai/mcp
userintuition-mcp login
userintuition-mcp list
userintuition-mcp list_studies

Start with a read such as list_studies to check your connection. See the current MCP setup guide for client-specific instructions and the research skills for workflow guidance.

Five workflow examples

1. Preference check

“Compare these three headlines with our intended audience. Ask me to choose panel or BYOP, create a draft, and help me review the study plan before recruitment.”

Claude uses create_study, customize_study, and get_study. A preference check uses the ordinary study workflow, not a separate quick-poll call.

2. Claim test

“Test what buyers think this claim means, whether they believe it, and what proof they would need.”

Include the exact wording and audience in the brief. Ask the agent to preserve skeptical reactions and to avoid turning a few interviews into a population-wide credibility score.

3. Message test

“Explore what this landing page promises and which parts are confusing. Help me supply the asset and approve the discussion guide.”

For concept tests, send an accessible concept image or URL through Customize Plan with its label and learning goals. The backend validates and attaches assets. Do not put raw study-plan fields or concept links into create_study.

4. Churn study with your own customers

“Use BYOP for customers who cancelled in the last 90 days. I will provide the authorized email list. Focus on the sequence leading to cancellation and alternatives considered.”

For a BYOP study, use create_participants with 1–100 unique participant emails per batch after the saved plan and invitations are approved. Invitations send by default; set silent: true on individual participant records when invitations should not send. Source customer lists through your own authorized export or integration; MCP has no direct CRM segment-sync tools.

5. Reuse prior evidence

“Compare these two studies, retrieve their reports, and check the interview sources behind the main differences.”

Agents can search findings and participant responses across authorized studies, then retrieve the underlying reports and interviews. Results preserve study context and source links; the calling agent interprets the evidence. Search coverage is explicit, and retrieving evidence does not launch research.

Common gotchas

A draft is not a plan. Create a metadata draft with create_study, then send the research brief through customize_study. Relay any planning questions to the user. Retrieve the persisted plan with get_study and obtain approval before recruitment. The user must choose panel or BYOP explicitly.

Review the estimate before launch. For a panel study, use launch_panel with dry_run: true to obtain the recruitment estimate. Show the country, language, cost, and timeline; launch with the same settings after approval. Each launch specifies one country. Audiences below 10% incidence require a feasibility request.

A transcript is not a completed study. Paginate interviews and check quality before counting them. Provisioning status reflects interviewer setup, not all fielding state.

Reports have a specific contract. Study results expose findings, participant responses, sample profiles, recommendations, and source references in JSON. Use generate_report when analysis is needed, and get_interview to verify supporting messages and recording links. Preference shares, credibility scores, and ranked themes are not guaranteed typed fields in this response.

Webhooks are supported. The server can create and delete completed-interview webhooks. They are account-wide and currently have no automatic delivery retries. Reconcile notifications against interview records when completeness matters.

A conversation is not a scheduler. Save the study ID and return to check progress, or use your own supported automation. Do not assume Claude will keep polling after its session ends.

Get started with a read

Ask Claude to list studies or inspect a selected study. When you are ready to commission research, provide the question and explicitly choose recruitment. See agentic research and the current tool reference.