An AI agent that can ask real customers questions is categorically more useful than one that can only reason about them. Getting there requires an interview API that supplies participants, runs moderated research-grade conversations, and returns structured findings — not just a voice synthesis layer the agent talks into.
This guide covers how the customer interview API pattern works, the three integration patterns agents commonly use, how recruitment operates programmatically, and how User Intuition’s MCP server differs from voice AI APIs like ElevenLabs. For full API documentation including tool schemas, see docs.userintuition.ai/mcp-server/overview.
What a customer interview API does, vs what survey APIs do
A survey API delivers questionnaires — structured prompts with fixed-choice or short-text responses. Survey APIs are fast, cheap per response, and statistically tractable at large n. They are the right tool when you need to measure: what percentage of users would use a feature, how satisfaction scores track over time, whether a respondent belongs to a behavioral segment.
A customer interview API does something different. It launches AI-moderated conversations that probe, follow up on unexpected answers, and surface the reasoning and mental models behind behavior. The output is not a distribution of choices — it is a corpus of explained positions that reveals why users behave the way they do and what they need from a solution that does not yet exist.
The tradeoff is cost and volume. Survey APIs work for large-n quantitative signals at low cost per response. Interview APIs work for understanding at the depth that changes product and positioning decisions. Both are part of a complete research stack; they are not substitutes.
For AI agents specifically, the relevant capability gap is this: a survey API cannot probe. When a respondent says “I prefer option B,” a survey records that. A moderated interview follows up: “What about option A didn’t land?” “What would option B need to do for it to be worth the switch?” The AI moderator extracts the reasoning, not just the choice. That reasoning is what changes decisions.
Three integration patterns
Pattern 1: A focused participant question
Use a narrow brief for a headline, claim, or concept. 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.
Pattern 2: A broader moderated study
For churn, discovery, or win/loss research, use the same draft and planning tools with a more detailed discussion guide. Choose an audience with direct experience of the decision. 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.
Pattern 3: Reuse selected study evidence
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.
Recruitment via API
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.
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.
How does User Intuition handle agent-driven customer interviews?
MCP exposes study planning, recruitment, interviews, results, evidence search, and supporting configuration operations. The CLI provides shell access to research workflows. Use the current tool catalog and API reference for exact names and arguments; tool counts vary by release.
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.
Use the current MCP reference for tool schemas and the public API reference for REST operations. The two surfaces share research resources but have different configuration contracts.
Comparison: voice AI APIs vs research-specific APIs
Voice AI platforms — ElevenLabs and similar platforms — provide real-time voice synthesis and conversation infrastructure. They are the right choice when the product is the conversation: customer support agents, voice assistants, interactive voice response systems. The API surface is optimized for real-time call handling: low-latency audio generation, call routing, transcript capture.
A research-specific API is built for a different use case. The conversation is a means to an end — the end being structured qualitative evidence. The distinctions that matter for research:
| Dimension | Voice AI API | Research API |
|---|---|---|
| Participant supply | None — you bring your own | 4M+ panel with targeting |
| Conversation design | Real-time, open-ended | Moderated research methodology |
| Output format | Raw transcript | Structured themes + verbatims + analysis |
| Recruitment | Not included | Screener, invite, incentive — all via API |
| Knowledge layer | None | Intelligence Hub for cross-study synthesis |
| Primary use case | Product is the conversation | Research produces evidence |
For agents that need to talk to customers the product already has, a voice AI API can be appropriate. For agents that need to recruit qualified strangers, conduct research-methodology conversations, and return analyzable evidence — the research-specific API is the correct layer.
Neither replaces the other. Some agent architectures use a voice AI layer for real-time customer support and a research API for periodic research programs. The distinction is: do you need participants, or do you already have them? Do you need research methodology, or is any conversation sufficient? Do you need structured findings, or is raw transcript enough?
Get an API key
For MCP integration (recommended for AI agents):
- Sign up at app.userintuition.ai/sign-up — Starter plan, free, three interviews on signup, no card.
- Generate a
ui_sk_key from Settings > API Keys. - Connect via stdio (
npx -y @userintuition-ai/mcp) or Streamable HTTP (https://mcp.userintuition.ai/mcp). - Full tool reference: docs.userintuition.ai/mcp-server/overview.
For a broader overview of the agentic research methodology — what agents can accomplish once connected and how the Intelligence Hub compounds over time — see agentic research.