MCP Server

Let your AI agent ask real people what they think — directly from a chat conversation. The User Intuition MCP server connects any MCP-compatible client (ChatGPT, Claude, Cursor, etc.) to your User Intuition account so you can create studies, collect feedback, and get results without leaving your AI workflow.

How It Works

The MCP server acts as a bridge between your AI agent and the User Intuition research platform.

AI Agent (ChatGPT, Claude, etc.)
    ↓  MCP Protocol
User Intuition MCP Server
    ↓  REST API
User Intuition Research Platform
    ↓
Real people respond via voice or text

Your agent gets access to five tools that cover the full study lifecycle — from creating a study to retrieving results.

Prerequisites

Before you begin, make sure you have:

A User Intuition account with credits in your wallet
An MCP-compatible AI client (ChatGPT, Claude Desktop, Cursor, etc.)

Connecting via ChatGPT

ChatGPT supports MCP servers natively. Connect User Intuition in a few clicks.

Open ChatGPT Settings

In ChatGPT, go to Settings → Connected Apps (or Tools)

Add MCP Server

Click Add MCP Server and enter the server URL:

https://mcp.userintuition.ai

Authenticate

ChatGPT will redirect you to sign in with your User Intuition account. Log in and grant access.

Start Using

Return to ChatGPT and ask it to run a study. For example:

“Ask 25 people which tagline they prefer: ‘Save time, save money’ vs ‘Less stress, more life‘“

Connecting via Claude Desktop

Add the MCP server to your Claude Desktop configuration.

Open Config File

Open your Claude Desktop MCP configuration file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json

Add the Server

Add the User Intuition MCP server to mcpServers:

{
  "mcpServers": {
    "userintuition": {
      "url": "https://mcp.userintuition.ai/mcp"
    }
  }
}

Restart Claude

Restart Claude Desktop. You’ll be prompted to authenticate with your User Intuition account on first use.

Available Tools

Once connected, your AI agent has access to five tools:

Tool	Description
ask_humans	Create a new research study and collect feedback from real people
get_results	Check study status and retrieve results
list_studies	List your studies with optional filters
edit_study	Modify a running study
cancel_study	Stop a running study

ask_humans

Create a new study to collect feedback from real people. This is the primary tool for launching research. Parameters:

Parameter	Required	Default	Description
`mode`	Yes	—	`"preference"`, `"claim"`, or `"message"`
`text`	Yes	—	The main content to test with respondents
`options`	Only for preference	—	2–5 options to compare
`context`	No	—	Background info shown to respondents
`audience`	No	`"general"`	Target audience segment
`emails`	No	—	Invite specific respondents by email
`screeners`	No	—	Screening questions to filter respondents (max 3)
`n`	No	`25`	Number of respondents (1–100)
`voice`	No	`false`	Collect audio responses
`time_budget_hours`	No	`3`	Max hours to keep the study open
`locale`	No	—	Language/region (e.g. `"en-US"`, `"es-MX"`)
`include_transcripts`	No	`false`	Include full response text in results
`redact_pii`	No	`true`	Auto-redact personally identifiable information
`dry_run`	No	`false`	Get cost estimate without creating study

Study Modes

Preference
Claim
Message

Compare 2–5 options to see which one people prefer.Great for testing taglines, names, designs, or any A/B decision.

"Ask 25 people which name they prefer for our app:
'Meadow', 'Bloom', or 'Canopy'"

Requires the options parameter with 2–5 items.

Test if people believe a claim. Find out whether a statement resonates or falls flat.Great for validating marketing claims, value propositions, or hypotheses.

"Test this claim with 30 people:
'Most parents would pay $20/month to save 3 hours on meal planning'"

Test message clarity. Understand what people think a message means and whether it’s clear.Great for testing copy, instructions, or communications before sending.

"Test this message with 20 people:
'Your plan will be upgraded automatically at the end of your trial'"

Screening Respondents

Use screeners to filter who responds to your study. Each screener is a multiple-choice question with pass logic that defines which answers qualify.

[
  {
    "question": "What is your current role?",
    "choices": [
      "Software Developer/Engineer",
      "Engineering Manager/Team Lead",
      "Non-technical role"
    ],
    "pass_logic": {
      "include": ["Software Developer/Engineer"]
    }
  }
]

Screeners are only available when audience is set to "general". You can add up to 3 screening questions per study.

Cost Estimates

Use dry_run: true to get a cost estimate before committing:

{
  "dry_run": true,
  "estimated_cost": 37.50,
  "currency": "USD",
  "eta_hours_range": [1.0, 3.0]
}

get_results

Check on a study and retrieve results when available. Parameters:

Parameter	Required	Default	Description
`study_id`	Yes	—	The study identifier
`include_transcripts`	No	`false`	Include full response transcripts

Study statuses: queued → fielding → complete Other statuses: partial, failed, cancelled When complete, results include preference splits, top reasons, minority objections, and verbatim quotes.

list_studies

List your studies with optional filters. Parameters:

Parameter	Required	Default	Description
`status`	No	—	Filter by status
`limit`	No	`10`	Max number of studies to return
`created_after`	No	—	ISO 8601 datetime filter

edit_study

Modify a study that is still in creating, queued, or fielding status. Only provided fields are updated.

Content changes (text, options, context, mode) will trigger moderation guide regeneration, which may affect in-progress interviews.

cancel_study

Stop a running study. Returns whether cancellation succeeded and the cost incurred for responses already collected.

Example Conversations

Here are some ways to use the MCP server through your AI agent:

Compare tagline options

You: “Ask 25 people which tagline they prefer for our fitness app: ‘Move more, stress less’ vs ‘Your body will thank you’ vs ‘Fitness that fits your life’”Agent: Creates a preference study, waits for results, and reports back:

52% preferred “Fitness that fits your life.” Top reason: it acknowledges that people have busy schedules. 24% chose “Move more, stress less” for its simplicity.

Validate a claim

You: “Test this claim with 30 parents: ‘I would switch grocery stores if one offered a 15-minute shopping guarantee’”Agent: Creates a claim study and returns:

71% agreed with the claim. Strongest agreement among parents with children under 5. Main objection: skepticism about freshness if shopping is rushed.

Test message clarity

You: “Test this email subject line with 20 people: ‘Your subscription changes take effect tomorrow’”Agent: Creates a message study and returns:

85% correctly understood this means their plan is changing. 15% were unsure whether they needed to take action. Suggestion: clarify whether action is required.

Dry run for cost estimate

You: “How much would it cost to survey 50 people about our pricing page?”Agent: Runs a dry run and reports:

A message study with 50 respondents would cost approximately $75.00 and complete in 1–3 hours.

Troubleshooting

Authentication Failed

If you see an authentication error:

Verify you’re logged into the correct User Intuition account
Try disconnecting and reconnecting the MCP server in your AI client
Ensure your account is active at app.userintuition.ai

Insufficient Credits

If a study fails with an insufficient credits error:

Use dry_run: true to check costs before creating studies
Add credits at app.userintuition.ai/wallet
The error message includes the exact amount needed

Study Not Completing

If a study stays in fielding status longer than expected:

Check progress with get_results — it shows completed vs. target responses
Studies with niche audiences or strict screeners may take longer
The time_budget_hours parameter controls the maximum duration

Tool Not Appearing in AI Client

If the User Intuition tools don’t appear:

Confirm the MCP server URL is correct: https://mcp.userintuition.ai
Restart your AI client after adding the server
Check that your client supports MCP (ChatGPT, Claude Desktop, Cursor)

Need Help?

Dashboard

View and manage your studies

Contact Support

Email our support team for assistance

Introduction

Getting Started

Creating a Study

Managing Your Study

Recruiting Participants

Integrations

Resources

How It Works

Prerequisites

Connecting via ChatGPT

Connecting via Claude Desktop

Available Tools

ask_humans

Study Modes

Screening Respondents

Cost Estimates

get_results

list_studies

edit_study

cancel_study

Example Conversations

Troubleshooting

Need Help?

Dashboard

Contact Support

Introduction

Getting Started

Creating a Study

Managing Your Study

Recruiting Participants

Integrations

Resources

​How It Works

​Prerequisites

​Connecting via ChatGPT

​Connecting via Claude Desktop

​Available Tools

​ask_humans

​Study Modes

​Screening Respondents

​Cost Estimates

​get_results

​list_studies

​edit_study

​cancel_study

​Example Conversations

​Troubleshooting

​Need Help?

Dashboard

Contact Support

How It Works

Prerequisites

Connecting via ChatGPT

Connecting via Claude Desktop

Available Tools

ask_humans

Study Modes

Screening Respondents

Cost Estimates

get_results

list_studies

edit_study

cancel_study

Example Conversations

Troubleshooting

Need Help?