The User Intuition MCP server exposes customer research as a structured API: 72 tools organized into 9 capability groups, two transport options, and a single API key. For developers asking “what can I programmatically call and how does it map to what a researcher actually does,” this guide covers the full surface. For teams evaluating agentic research workflows, the 9-group structure maps directly to research workflow phases — so understanding the groups tells you which tools belong at each step.
Full schema-level documentation, including parameter types and response shapes, lives at docs.userintuition.ai/mcp-server/overview.
What an MCP customer research API exposes
A conventional research platform exposes its functionality through a web dashboard. An MCP customer research API exposes that same functionality as typed, callable tools. The distinction matters because it changes who (or what) can use it.
A dashboard requires a human researcher to navigate, configure, and interpret. An API lets software programs and AI agents perform the same operations. The research workflow — configure a study, recruit participants, run interviews, retrieve findings — becomes a sequence of function calls with typed inputs and structured outputs.
The MCP protocol adds a discovery layer: an agent connects to the server and receives a manifest of all available tools with their schemas. The agent can then call any tool using natural language through a compatible AI client (Claude, ChatGPT, Cursor) without bespoke integration code. The same connection pattern works across every MCP-compatible host.
The User Intuition MCP server implements this for the full research workflow. The server is published to npm as @userintuition-ai/mcp v0.2.5. The 72 tools are organized into 9 groups that reflect distinct phases of the research lifecycle.
The 9 capability groups
Human Signal — 5 tools
ask_humans, get_results, list_studies, edit_study, cancel_study
Human Signal is the direct path to panel signal without full study configuration. ask_humans dispatches a question to a panel segment and returns structured responses, typically within hours. It is optimized for directional checks — headline preference, claim credibility, option preference — where a full moderated study would take longer than the decision cycle allows.
get_results retrieves responses once fielding completes. list_studies surfaces all studies in the workspace including their current status. edit_study modifies a live study’s parameters; cancel_study terminates fielding early. The group is the entry point for agents that need real signal fast without standing up the full moderated-study pipeline.
Studies — 13 tools
list_assistants, get_assistant, create_assistant, update_assistant, delete_assistant, create_assistant_from_invite_id, stop_panels, upload_concept_link, delete_concept_link, list_available_panel_questions, list_available_languages, list_available_modes, list_study_types
The Studies group handles the AI-moderated interview study lifecycle. create_assistant is the primary creation tool — it accepts study type, discussion guide, audience configuration, interview modality, and language, and returns a study ready for participant recruitment.
Important operational note: update_assistant requires a prior get_assistant read to function correctly — it is not idempotent on partial fields. screener_questions on update replaces the existing screener entirely rather than merging; pass the complete intended screener list. Studies are not automatically created with screeners — screener configuration is a separate step.
list_available_languages surfaces 50+ supported languages; list_available_modes returns available modalities (audio, chat, video); list_study_types returns configured study templates. upload_concept_link and delete_concept_link handle stimulus materials for concept tests.
Invites & Participants — 8 tools
list_invites, get_invite, create_invite, update_invite, delete_invite, mark_invite_sent, bulk_create_invites_from_segment, send_reward
This group manages participant recruitment. bulk_create_invites_from_segment is the high-leverage call — it creates invitations for an entire audience segment in a single operation, applying panel targeting filters (role, industry, company size, seniority, behavioral attributes, language, geography) with study configuration. The 4M+ panel covers B2B and B2C segments across 50+ languages and geographies.
For bring-your-own-participant studies, create_invite handles individual invitations. send_reward distributes participant incentives once interviews complete. mark_invite_sent records that a custom invitation was dispatched through an external channel.
Calls & Interviews — 7 tools
list_calls, get_call, update_call, delete_call, get_call_usage_stats, get_calls_grouped_by_invite, analyze_transcripts
The Calls and Interviews group provides access to interview records and analysis. list_calls and get_call surface the raw interview data. analyze_transcripts is the primary analysis tool — it takes a set of call IDs and returns structured findings: recurring themes, key quotes, sentiment patterns, and cross-participant comparisons.
get_calls_grouped_by_invite organizes call records by the invite that generated them, which is useful for comparing responses across audience segments when multiple invite batches were created for the same study. get_call_usage_stats provides credit consumption data per call.
Voice & Reports — 2 tools
list_vapi_voices, get_assistant_report
Voice and Reports handles voice configuration and individual study reporting. list_vapi_voices returns available voice options for audio-mode studies. get_assistant_report retrieves the formatted report for a completed study — the structured summary of findings, themes, and recommendations generated from all interviews in the study.
Intelligence Hub — 18 tools
get_file_search_store, list_documents, delete_document, query_intelligence, list_chat_history, create_session, list_sessions, get_session, update_session, patch_session, delete_session, create_studio_output, list_studio_outputs, get_studio_output, update_studio_output, delete_studio_output, generate_report, generate_powerpoint
The Intelligence Hub is the largest group and the most important for agents that run recurring research. query_intelligence runs a natural-language query over all completed studies in the workspace and returns synthesized findings with source references. It draws on transcript-level detail, not summaries, so answers reflect full interview depth.
generate_report and generate_powerpoint produce formatted deliverables. Session management tools (create_session, get_session, update_session, patch_session, delete_session, list_sessions) support multi-turn research conversations. Studio output tools handle saved analysis artifacts that can be retrieved and updated across sessions.
The hub compounds in value: a workspace with 50 completed studies answers cross-study questions far more reliably than one with 5. Agents that run recurring research pipelines should treat the Intelligence Hub as the accumulating institutional memory of the workspace.
Integrations & Panels — 5 tools
list_segments, list_external_participants, get_integration_status, trigger_integration_sync, create_panel_survey
The Integrations and Panels group handles external participant sources and CRM connections. list_segments returns defined audience segments available for targeting. list_external_participants surfaces participants imported from connected CRMs or external panels. trigger_integration_sync refreshes participant data from a connected integration. create_panel_survey creates screener surveys for custom panel qualification. get_integration_status reports the health of configured integrations.
Monetization & Utilities — 8 tools
list_transactions, get_transaction_receipt, list_plans, get_plan, redeem_coupon, list_referrals, send_referral_invite, resend_referral_invite
Monetization and Utilities covers billing, credit management, and referral operations. list_transactions and get_transaction_receipt surface credit consumption history. list_plans and get_plan expose available subscription options. redeem_coupon applies promotional credits. The referral tools handle the referral program workflow.
Account — 6 tools
get_my_organization, list_org_members, update_my_profile, submit_feedback, contact_sales, contact_support
Account tools manage workspace and profile state. get_my_organization retrieves workspace configuration and current plan status — useful for agents that need to check credit balance before launching a study. list_org_members surfaces team members with their roles. update_my_profile modifies user profile fields. The contact tools (submit_feedback, contact_sales, contact_support) create support and sales requests programmatically.
Three workflow patterns
The 9 groups support three distinct research patterns matched to different decision velocities.
Pattern 1: Rapid panel poll (Human Signal)
For directional signal within hours without full study configuration:
ask_humans(question="...", segment="...", sample_size=20)
→ [poll in progress]
→ get_results(study_id=...)
→ Returns: preference split + representative explanationsEntry group: Human Signal. Decision velocity: same day.
Pattern 2: Full moderated study
For research-depth findings on 10-50+ participants over 24-48 hours:
create_assistant(study_type="...", discussion_guide=[...], language="en")
→ bulk_create_invites_from_segment(assistant_id=..., segment_filters={...})
→ [interviews complete via AI moderator]
→ analyze_transcripts(call_ids=[...])
→ generate_report(assistant_id=...)Entry groups: Studies → Invites and Participants → Calls and Interviews → Voice and Reports. Decision velocity: 24-48 hours for standard studies.
Pattern 3: Intelligence Hub query
For synthesized answers over accumulated research in seconds:
query_intelligence(query="What objections surface most in enterprise pricing conversations?")
→ Returns: synthesized findings with source references across all completed studiesEntry group: Intelligence Hub. Decision velocity: seconds when prior research exists. This pattern requires an established workspace with completed studies; value compounds with each study added.
Authentication and transport selection
Authentication: Set USERINTUITION_API_KEY to your ui_sk_ key. No scopes, no per-group tokens, no webhook configuration required. The key covers all 9 groups. Multiple keys can coexist per workspace; any key can be revoked without affecting others.
Transport — stdio (local):
npx -y @userintuition-ai/mcpRuns as a local subprocess. Suited for development, IDE integrations (Cursor, VS Code with MCP extensions), and local Claude Desktop configuration. Requires USERINTUITION_API_KEY set in the environment before starting.
Transport — Streamable HTTP (cloud):
Endpoint: https://mcp.userintuition.ai/mcp
OAuth-authenticated. Suited for production agent deployments, cloud-hosted AI pipelines, and CI workflows where running a local subprocess is not feasible. The same ui_sk_ key is exchanged via OAuth at connection time.
Rate limiting: The API enforces rate limits at the workspace level. get_call_usage_stats and list_transactions in the Calls and Monetization groups provide consumption data. get_my_organization in the Account group returns current plan limits.
How does User Intuition handle a programmatic customer research API?
User Intuition built its MCP server as the primary integration surface for AI-agent customer research, not as a secondary developer feature. The 72-tool surface covers the entire research workflow as first-class API operations — nothing is reserved for dashboard-only access.
The 9-group structure reflects deliberate API design: Human Signal for speed, Studies for configuration, Invites and Participants for recruitment, Calls and Interviews for retrieval, Intelligence Hub for synthesis. An agent that understands the groups understands the research lifecycle.
Three design choices make the API production-ready for agent workflows. First, authentication is a single ui_sk_ key — no scopes, no multi-token choreography, no webhook registration required. Second, both transports are first-class: stdio for development, Streamable HTTP at https://mcp.userintuition.ai/mcp for cloud deployments. Third, the Intelligence Hub’s query_intelligence tool turns the API from a transaction layer into a compounding knowledge layer — each study added to the workspace increases the quality of answers to every subsequent cross-study query.
The server ships on npm as @userintuition-ai/mcp v0.2.5 and is accessible via any MCP-compatible host: Claude, ChatGPT, Cursor, or custom agent frameworks built on the MCP SDK. The 4M+ participant panel, 50+ language coverage, and AI-moderated audio/chat/video interview modes are available through the same tools that cover the rest of the workflow. There is no separate recruitment API, no separate analysis API, and no separate intelligence API — the 9 groups are one coherent surface. Full documentation at /platform/agentic-research/.
Get started
Full schema documentation including parameter types, response shapes, and worked examples: docs.userintuition.ai/mcp-server/overview.
To connect via stdio immediately:
npx -y @userintuition-ai/mcpSet USERINTUITION_API_KEY to your ui_sk_ key. No card required on the Starter plan — three free interviews included on signup, with the full 72-tool API surface available from day one.
The Starter plan is free with pay-per-use credits ($25/credit for audio, $12.50 for chat). The Professional plan ($999/month) includes 50 credits monthly and reduces the per-credit rate to $20 for audio ($10 for chat). Plan and credit details available via list_plans in the Monetization group once connected, or at docs.userintuition.ai/mcp-server/overview.