Announcing User Intuition MCP: 72 Tools for Customer Research

We just released the User Intuition MCP server — 72 tools, 9 capability groups, two transports, no dashboard required.

The agentic research platform now has a complete programmatic surface. Any AI agent that supports the Model Context Protocol can connect, launch real customer studies, recruit from the 4M+ panel, and retrieve analyzed findings — all without a human opening the dashboard.

What Just Shipped?

@userintuition-ai/mcp v0.2.5 on npm. One install line:

npx -y @userintuition-ai/mcp

Spec summary:

• 72 tools across 9 capability groups
• 4 built-in prompts for common research workflows
• Two transports: stdio (Claude Desktop, Cursor, VS Code) and Streamable HTTP with OAuth (hosted agents)
• One key format: ui_sk_… — set as USERINTUITION_API_KEY env var

The canonical reference is at docs.userintuition.ai/mcp-server/overview.

The 9 Capability Groups

1. Human Signal (5 tools)

The core research loop: ask_humans, get_results, list_studies, edit_study, cancel_study. This is where an agent turns a research question into a live study with real participants and retrieves structured qualitative findings. ask_humans supports preference, claim, message, and several other modes.

2. Studies (13 tools)

Full assistant lifecycle management: create_assistant, update_assistant, delete_assistant, get_assistant, list_assistants. Plus study configuration tools: list_available_modes, list_available_languages, list_study_types, list_available_panel_questions. For more complex flows: create_assistant_from_invite_id, stop_panels, upload_concept_link, delete_concept_link.

3. Invites & Participants (8 tools)

Manage who gets interviewed: create_invite, list_invites, get_invite, update_invite, delete_invite, mark_invite_sent. The high-leverage tool here is bulk_create_invites_from_segment — connect your Shopify or HubSpot customer list and recruit from a real segment. Also: send_reward for participant compensation.

4. Calls & Interviews (7 tools)

Monitor and work with completed conversations: list_calls, get_call, update_call, delete_call, get_call_usage_stats, get_calls_grouped_by_invite. The key analytical tool is analyze_transcripts — runs structured theme extraction and sentiment analysis across a study’s call set.

5. Voice & Reports (2 tools)

list_vapi_voices and get_assistant_report. The report tool returns the full synthesized findings document for a completed assistant — themes, evidence, recommendations — in structured form.

6. Intelligence Hub (18 tools)

The compounding layer. query_intelligence lets an agent query across all past studies like a knowledge base. generate_report and generate_powerpoint produce formatted deliverables. Full session management (create_session, list_sessions, get_session, update_session, patch_session, delete_session), Studio outputs for saved artifacts, document management (list_documents, delete_document), and list_chat_history.

7. Integrations & Panels (5 tools)

list_segments, list_external_participants, get_integration_status, trigger_integration_sync, create_panel_survey. This is how agents access CRM-connected customer lists and check sync health.

8. Monetization & Utilities (8 tools)

Account credits, transactions, plans, coupons, and referrals: list_transactions, get_transaction_receipt, list_plans, get_plan, redeem_coupon, list_referrals, send_referral_invite, resend_referral_invite.

9. Account (6 tools)

Org and member management, plus support channels: get_my_organization, list_org_members, update_my_profile, submit_feedback, contact_sales, contact_support.

How to Get Started

Step 1: Get an API key. Sign up at app.userintuition.ai/sign-up — Starter plan is free, 3 interviews, no credit card.

Step 2: Add to your agent config.

For Claude Desktop (add to ~/Library/Application Support/Claude/claude_desktop_config.json):

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

For Streamable HTTP / hosted agents, point at https://mcp.userintuition.ai/mcp with OAuth.

Step 3: Test the connection. Ask your agent: “List my User Intuition studies.” It calls list_studies and returns your study history. If you have no studies yet, it returns an empty array — which is correct.

Step 4: Run your first study. Ask: “Run a preference check on these 3 tagline options with 25 people.” The agent calls ask_humans with mode: "preference", recruits participants, and returns results in 2-3 hours.

Why Agent-Native Matters

The dashboard was always one way to run research. It was never the only way. The MCP server makes AI agents a peer client — same studies, same panel, same Intelligence Hub — just accessed programmatically.

The practical implication: research stops being a manually-scheduled event and becomes an automated step in any agent workflow. A product agent can validate a feature assumption before writing a spec. A marketing agent can test a claim before generating campaign copy. A strategy agent can check competitive perception before finalizing a quarterly plan.

Real human signal — not synthetic, not scraped, not modeled — as a routine step in an automated workflow. That is what 72 tools and two transports enable.

The dashboard still works exactly as before. Existing users keep every study, every finding, every Intelligence Hub query. The MCP server adds a new access path; it does not change anything about the existing one.

How Does User Intuition Handle Agent-Native Research?

User Intuition’s agentic research platform was built from the ground up for programmatic access, not retrofitted. The 72-tool MCP surface reflects the full depth of the platform — not a subset API bolted on afterward. When an agent calls create_assistant, it creates the same AI-moderated interview assistant that the dashboard creates. When it calls analyze_transcripts, it runs the same NLP pipeline. When it calls query_intelligence, it queries the same Intelligence Hub that stores all your past research.

The panel is real: 4M+ vetted participants across 50+ languages, with multi-layer fraud prevention (bot detection, duplicate suppression, professional respondent filtering). Studies return in 24-48 hours. Cost is $20 per audio interview on the Pro plan. Participant satisfaction averages 98% — meaning the conversations are genuine and engaging, not checkbox completions.

The Intelligence Hub compounds across every study run by any client — dashboard, Claude, Cursor, or a custom agent. An agent calling query_intelligence today gets access to findings from studies you ran six months ago via the dashboard. The knowledge base is unified regardless of how the research was run. This is what makes the 72-tool surface more than a convenience: it connects agent-driven research to institutional memory that grows with every study.

Get an API key at app.userintuition.ai/sign-up, then npx -y @userintuition-ai/mcp and you are done.