Agentic Research

Qualitative research for AI agents, through MCP and API

Your agent turns a decision into a study, recruits real participants, runs AI-moderated interviews and brings back findings cited to the moment each participant said it. From Claude, ChatGPT, Codex or your own code.

Works in Claude, ChatGPT, Codex, Cursor & VS Code
You set the spending rules: approve each launch or set a monthly budget

Install from: Claude plugin · ChatGPT plugin · Cursor directory

Researcher using User Intuition AI-moderated research platform
Research workspace Preview
5–7Levels of Laddering
4M+Vetted Panelists
98%Participant Satisfaction
Fieldwork activity7 days
One place for your evidence

Calls, transcripts, and findings.

Trusted by teams at

Capital One
RudderStack
Nivella Health
Turning Point Brands
Procter & Gamble
Microsoft
CHG Healthcare
TL;DR

A research API and MCP server lets an AI agent commission interviews with real people and bring back evidence for a decision. With User Intuition, the agent turns a brief into a saved plan and cost estimate that you approve before launch. Its AI moderator runs 20–30 minute interviews that ladder 5–7 levels deep, with participants from a vetted 4M+ consumer panel or your own users. Findings come back with prevalence and verbatim quotes linked to each source interview, in Claude, ChatGPT, Codex, Cursor or over REST, at $30 per voice interview.

At a glance

What is a research API for AI agents?

User Intuition is the research API and MCP server for AI agents: through API, MCP and CLI, your agent turns a decision into a study brief, launches recruitment once you approve the plan, and retrieves findings linked to real interviews. Each 20–30 minute AI-moderated interview ladders 5–7 levels deep, with most panel studies filling within 24 hours from a vetted 4M+ consumer panel across 58 countries, or your own customer list, run in parallel as participants respond.

Who uses the research API and MCP server for AI agents?

Research agencies and in-house research teams running studies from Claude, ChatGPT, Codex or an IDE, and builders whose agents need real participant evidence before they act.

How does an AI agent run qualitative research?

The agent drafts a study plan from your brief, you approve the saved plan and recruitment estimate, and the AI moderator interviews participants in voice, video or chat. Results come back as structured JSON with each finding tied to its source interviews.

What does your agent get back from a study?

Study findings with prevalence, participant responses, participant profiles and recommended next steps, plus the verbatim quotes and source interviews your agent can cite.

The problem, and how we fix it

Why does an AI agent need real interviews for research?

An agent can draft hypotheses all day; a decision still needs evidence from real people. Here's what builders and research teams tell us, and how User Intuition closes each gap.

  1. Sound familiar?

    Our agent can't run a study through a browser. It clicks around, then someone exports a CSV by hand.

    Documented API, MCP and CLI operations

    Create and customize studies, estimate recruitment, run interviews and retrieve results through documented operations with explicit inputs and saved state. You review the plan and estimate, then approve each paid launch or set a monthly budget the agent launches within; read the API documentation or start at the agent homepage.

  2. Sound familiar?

    A survey API tells us people picked Option A. It can't ask them why.

    AI-moderated interviews, not static surveys

    The moderator ladders 5–7 levels from a stated preference to the decision driver beneath it, in voice, chat or video, and surfaces the minority objection a survey never reaches.

  3. Sound familiar?

    Scraped reviews tell us what people said last year, not how they react to our message today.

    Fresh reactions from a defined audience

    Recruit from a vetted 4M+ consumer panel across 58 countries, with B2B audiences recruited to order per study, or invite your own customers, and review the cost and timing estimate for your sample, audience and country before approving paid recruitment.

  4. Sound familiar?

    The model predicted how customers would react. Nobody can tell which parts came from a real person.

    Real participant evidence, separate from interpretation

    Every finding carries source references to the interview behind it, and your agent can search prior findings and participant responses before commissioning more research. Run the public examples to see the workflow.

Developer Setup

Connect Your AI Agent in Minutes

Connect once, then move from a research brief to approved recruitment and source-backed findings.

1
Connect

Authorize Your Client

Add User Intuition Research in Claude, a developer-mode connector in ChatGPT, or the hosted server in Codex and IDEs. REST clients use an API key.

2
Design

Customize the Study

Choose your panel, your own participants or synthetic respondents. Send your brief, audience and concept material through the planning step, then review the saved study.

3
Recruit

Recruit From the Vetted Panel

Target the audience by country, demographics and screener from the 4M+ vetted consumer panel, or invite your own participants. B2B audiences are recruited to order per study.

4
Approve

Approve the Launch or Set a Budget

The agent returns the cost and timeline estimate alongside the saved plan. A person approves the quote: one click in Claude, a confirmation in the conversation elsewhere. Or a workspace owner sets a monthly panel budget, and the agent launches one-time panels within it on its own.

5
Field

Real Interviews Run

Participants join the AI-moderated interview. Your agent tracks progress and quality as interviews complete, and your team can watch any interview live.

6
Analyze

Inspect the Findings

Retrieve findings with verbatim quotes and the supporting interviews, then ask follow-up questions across your research library to build the recommendation.

Methodology & Trust

Choose a Focused Question or a Broader Study

Use a focused brief for messaging or concept questions, or a broader guide for exploration. Choose the design around the decision, audience, and evidence you need.

Messaging & Concept Research

  • Understand how people interpret a headline or tagline
  • Explore why one concept feels clearer or more relevant
  • Probe whether a claim is believable and why
  • Identify objections before revising a message
  • Test a focused change with a defined audience
  • Review the estimated recruitment cost and timeline

Broader Research Questions

  • Explore purchase decisions, churn, and unmet needs
  • Review audience qualifications and custom screening
  • Test a concept image or a live prototype link
  • Inspect reports alongside the supporting interviews
  • Compare evidence across studies and audiences
  • Invite a supplied list of customers or recruit a panel

Both approaches use a reviewed study plan. Retrieve findings, participant responses and supporting interviews, or search every past study and get answers cited to the quote.

Why User Intuition

Why researchers choose User Intuition

User Intuition is an AI-moderated interview platform that gives research teams and agencies the depth of a senior researcher at the scale of a survey, with vetted respondents, your own methodology and evidence you can trace to the verbatim. Quality is built in: a person vets every panelist by hand, every session is screened for fraud, and you pay only for interviews that pass.

Senior-researcher depth

The moderator ladders 5–7 levels deep by default, using Reynolds and Gutman's laddering method, so you hear the reasons behind the reasons.

5–7 levels by default

Your methodology, run as written

Bring your discussion guide and frameworks. Required questions stay pinned in order, word for word if you need it, study rules keep the moderator inside your compliance lines, screeners and hard quotas fill every cell to plan, and you decide how far it probes.

Your guide, your probing

Every finding traces to its source

Click any theme through to the verbatim, the transcript and the recording. Every claim in the deliverable traces to what a participant said.

Quote → transcript → recording

Respondents you can stand behind

A 4M+ vetted panel across 58 countries: our team reviews a video identity interview with every panelist, every voice interview is screened for AI-generated or coached answers, and panelists are tracked across studies. Or bring your own customers, panel provider or survey respondents. Either way, you can watch any interview live.

Vetted by hand, tracked over time

Deliverables ready to present

Themes, verbatims and an editable presentation for every study, plus every transcript, recording and screener answer to export into your own tools. Put your own brand on them when you need to.

Editable deck, report, transcripts

Survey scale, quality-only billing

200+ panel interviews in 24 hours, with no cap on parallel conversations. Each one is scored on length, depth and coverage, and you pay only for those that pass.

200+ panel interviews in 24 hours
  • Isolated workspaces
  • Confidential stimulus
  • Never used to train AI models
  • GDPR & CCPA compliant
  • SOC 2 Type II compliant
  • Security →
Compare

Compare the Research Workflow

Dimension User Intuition MCP Dashboard Research Workflows Data & Scraping APIs
Agent access MCP research tools, plus REST API and CLI Product-specific: check API or connector availability Programmatic search and extraction
Interview depth AI-moderated conversations with 5–7 level laddering Moderation capabilities vary by platform No interviews — static data or social content
Real people Yes — vetted 4M+ consumer panel or your own list Participant sources and export options vary Depends — some panels exist, no moderation
Result format Findings with prevalence, verbatim quotes and source interviews in JSON Reports, exports, and APIs depend on the product Documents, excerpts, and structured outputs vary by API
Fresh signal New interviews after plan and recruitment approval Setup and automation vary by platform Historical or batch — not specific to your content today
Cost $30 per quality voice interview, or $60 with standard panel recruitment; preview the total Check the platform and recruitment quote Usage-based pricing varies by service
Prior evidence Ask questions across your research library with cited answers Repository and connector features vary Public-source retrieval; private research requires a separate source
Sample Research

Hear the interviews. See the presentation.

Explore calls and a sample presentation from our 43-participant Walmart shopper study.

W10
10 min

Grocery top-up, supercenter

“Basically, because we use those things all a lot We try to always have you know, certain things in the house at all times because we use them very frequently. So we don't like to run out of them.”

Routine top-upOne-stop shopperPrice aware
W30
14 min

Phone replacement, unplanned and urgent

“It made me feel so safe so relieved, and it brightened my day.”

Urgent replacementStaff-assistedNew to the area

Sample presentation

Preview three slides, or download the full 15-slide presentation.

Title slide: How Walmart shoppers decide. 43 shoppers, 7.6 hours of interview, 34 who hit an obstacle, 36 who will return anyway.

1 / 3How Walmart shoppers decide

Download PDF

Bring your own sample

$30 per quality voice interview

Use our 4M+ vetted panel

$60 per quality voice interview

Customer stories

What researchers say

"It's quick to set up, the AI interviews actually hold up, and everything we've ever learned is searchable in one place. That's the direction consumer insights should have gone years ago."
Aldrin M., Operations Lead, Computer Software Verified review on Capterra
"I switched from SurveyMonkey to User Intuition because it allows me to collect real-time voice data from interviews, which is key for me. The initial setup was not hard at all; Charles, the AI, helped me set it up by using my parameters, and it generated everything I needed — just a bit of editing was required. It fits perfectly into our consulting workflow, as we don't do many interviews and it's cost-effective for our small team."
Willys N., Consulting, Small Business Verified review on G2
FAQs

Frequently Asked Questions

In Claude, add User Intuition Research from the connector directory and sign in. In ChatGPT, add a developer-mode connector pointing at https://mcp.userintuition.ai/mcp; OAuth is discovered automatically. Codex, Cursor, Claude Code and VS Code connect to the same hosted server or run @userintuition-ai/mcp locally with an API key.

Claude, ChatGPT, Codex, Cursor, Claude Code, VS Code and any MCP-compatible agent. Developers can also call the REST API directly or use the CLI from scripts and CI.

Yes. Sign in through OAuth or create an API key. A free account includes 3 free interviews with your own participants; panel recruiting is billed separately.

Yes. Planning, recruitment, fielding, monitoring, reports and research questions all run through MCP or the REST API, and the CLI covers study and results workflows from scripts. The dashboard is there when you want to listen to interviews, watch one live or share results visually.

Agent-run studies use the same pricing as the dashboard: $30 per voice interview with no subscription, or $60 with standard panel recruitment. The agent shows you the full estimate before a launch, and you can set a monthly budget that caps what it spends.

Yes. You pay only for interviews that pass our length, depth, and coverage checks, whether a study runs from the dashboard or through an agent.

Yes. Ask the agent for a concept or message test with an image or prototype link and a defined audience. It drafts the study, shows you the saved plan and estimate, and returns findings on what participants understood, preferred and why.

Yes. Choose your own participants and the agent invites them by email, or share the study link anywhere your users are. A separate study can recruit from the panel, or use synthetic respondents in chat for a fast directional read.

Structured JSON: study findings with prevalence and verbatim quotes, participant responses, participant profiles and recommended next steps, each tied to its source interviews, with recording timestamps in voice and video studies. Webhooks notify your system when interviews and reports complete.

Agentic research is research an AI agent plans, launches and analyzes on your behalf, with real participants. With User Intuition, the agent designs the study from your brief, recruits after you approve the plan and estimate, and returns evidence cited to the quote.

A consumer research API lets software create research studies, recruit participants and retrieve their evidence. User Intuition's API, MCP server, CLI and skills cover study design, recruitment, AI-moderated interviews, cited reports and cross-study search.

Yes. The agent can ask a question across your research library in plain language and get an answer cited to the source, then open the reports and interviews behind it.

The public examples repository includes TypeScript examples for running a study, retrieving results and searching evidence, with fixture mode so you can try the flow before connecting an account.
Get Started

Give Your Agent Real Participant Evidence

Connect your account and explore the research tools, CLI, and skills.

Start Free

Sign up, generate an API key, and run your first study.

MCP Server Overview

Installation, transports, tool reference, and quickstart guides.

Staging Key

Test your agent's study workflow against the staging API before you run a live study.

Works with Claude, ChatGPT, Cursor, Claude Code, VS Code, and any MCP-compatible agent.

Explore More

Related resources

Alternatives & Comparisons

Side-by-side comparisons with competing platforms and approaches.

Related Solutions

Complementary research use cases that pair with this topic.

Industries

See how teams in specific verticals apply this research.

Last updated