User Intuition

Legacy bulk selection: add every currently available study to this Hub session. Use sync_research_study for a selected study.

Answer a question across selected accessible studies using indexed research. Returns generated statements with numbered citations to canonical content IDs, plus source details and index status. Use search_research to inspect the underlying records.

MCP App button action. Consumes one short-lived approval challenge, rechecks the saved study and price, then launches one paid panel cycle.

Withdraw a pending feasibility request. A request already answered cannot be cancelled.

Cancel one durable analysis run only when the user requests it. Existing results remain available for inspection.

Cancel a queued or running Hub export. A cancelled worker cannot publish its file.

Wait for a chat answer; if still running, return the same durable job without cancelling it. For study findings, feedback, comparisons and quotations, use this tool with primary File Search and verified quotations. Resolve ambiguous study names first. Reuse the selected Hub session; if absent create one with create_research_session and a stable retry key, inspect sources and add only missing requested studies. Do not resync unchanged sources or broaden the session silently. After completion always call view_research_answer to display the answer and clickable quote/recording cards inside this conversation, without redirecting to a dashboard; preserve the answer and citation labels. Poll inspect_research_session view=job. Set parent_job_id for a follow-up in the same session. Reuse the same idempotency_key and prompt after a timeout; sync and async share the job.

Legacy bulk removal: remove all sources from this Hub session without deleting the underlying studies or files. Use only when the user requests clearing the session.

Create or replace a customer-supplied panel bridge on an owned BYOP study. Supply completion and screen-out redirect URLs; set the provider's respondent ID query parameter and placeholder if the defaults differ. Optional session, external study, quota-full, and technical-failure mappings are supported. Before sharing entry_url with the provider, call get_study, verify provisioning_status=provisioned, and have the human approve the current plan and URL mapping. This does not recruit respondents or launch the paid User Intuition Panel. Repeating the same PUT preserves the current entry token.

Queue creation of 1–100 unique BYOP participants. Poll get_participant_job until succeeded, then list participants. Before creating participants, call get_study, require provisioning_status=provisioned, and verify or obtain explicit approval of the current plan. Carry a trusted unchanged approval forward using study_id and updated_at; changed plan or invitation settings invalidate it. Invitations send unless silent is true.

Create an organization Hub session for selected studies and uploaded sources. Reuse the same idempotency key after an uncertain response.

Create an in-depth-interview, concept-test, or prototype-test draft using ordinary metadata only. Ask the human to explicitly choose recruiting_method (panel, BYOP, or synthetic respondents). A value proposed or inserted by a host/delegating model is not human confirmation unless it quotes the human's actual choice. Unless the human requests otherwise, use a voice interview in English with Elliot; these defaults are applied when interview_format, language, or voice are omitted. Synthetic respondents require chat. A prototype-test must use video or voice, never chat. Do not draft or pass the study plan, audience targeting, screener questions, concept links, or concept images here; after creation, send the user's natural-language research brief to customize_study. For a concept test, pass an accessible attachment through customize_study.concept_image or include a publicly downloadable concept-image URL in the message, plus its participant-facing label, stimulus context, audience, and learning goals. For a prototype test, include the prototype URL, participant-facing label, intended tasks, audience, and learning goals. The Customize Plan backend validates and attaches the asset. Voice configuration is kept for chat as well as voice and video.

Create a webhook for an account or one owned study. Choose interview, study lifecycle, participant lifecycle, or quota events. The signing secret is shown once; store it securely.

Send one research-design turn to Customize Plan. This operation may fetch a researcher-provided concept URL and use an external analysis model. A 2026 Tasks client receives a durable task for turns without concept_image; poll tasks/get for the completed planning turn. Native concept images use the synchronous call. Put operational workflow constraints in execution_policy, never in message. Set decisions=agent only when the human delegated research-design choices; report assumptions and relay only remaining required questions. Otherwise use human mode and relay questions[]. If the client exposes an attached image as bytes or a readable local path, ask for a participant-facing label when missing, base64-encode the raw bytes, and send concept_image. Completed plans and setting changes are persisted. Then get_study, return the full saved plan for review, and obtain approval before invitations or paid launch.

Disable the external-panel entry link for an owned BYOP study. Existing interviews remain; the provider can no longer send new respondents through this link. A repeated deletion returns 404.

Soft-delete an interview owned by the authenticated user. Call only when the human explicitly requests deletion of this interview.

Soft-delete one BYOP invitation after the user explicitly requests it. Deletion also removes its personal link. If interviews exist, the request returns a conflict unless the user separately approves deleting those interviews and cascade_interviews is true.

Permanently delete an existing Studio output in its session only when the user requests deletion. Read the output first.

Delete one organization Hub session only when the user explicitly requests it.

Soft-delete a study and its linked invitations, interviews, and reports from normal reads. Participant links stop working. There is no public restore endpoint. Uploaded assets may have separate retention. Call only when the human explicitly requests deletion of this study.

Deprecated compatibility tool. List webhooks, then use delete_webhook_by_id so deletion works through clients that omit DELETE request bodies. A repeated deletion returns 404.

Stop future deliveries to one registered webhook. A repeated deletion returns 404.

Retrieve a saved report as Markdown, PowerPoint as PPTX/PDF, or Highlights as a short-lived private MP4 link. Files return filename, mime_type and data_base64 (up to 64 MB); decode to an artifact rather than showing base64 to the user. Access is checked again. Preserve verified references; never invent download URLs.

Apply the requested prompt to an existing report or PowerPoint in place. Read the output first. Retains its ID and reuses saved evidence; default reports and Highlights cannot be edited. Poll inspect_research_session view=job and read the original output after success. Reuse the key on retry.

Read-only estimate for one Panel recruitment cycle. Returns per-interview and total credit cost, available balance and deficit, participant reward, card hold, recurrence, and a heuristic timeline. launch_panel shows this estimate for human App approval, or human chat confirmation in clients without Apps, before the paid request. The price and balance can change before launch; use submit_feasibility_request for incidence below 10% or more than 1,000 interviews.

Queue analysis of study responses using an external analysis model. A 2026 Tasks client receives a durable task and polls tasks/get for the completed job; other clients receive a job ID and poll get_report_job. Then call get_study_report. report.ready webhooks also signal completion.

Queue a report, native PowerPoint or saved Highlights video in an organization Hub session. Reuses saved reports and findings; no need to run exhaustive analysis first. Highlights follow default-report learning goals/quotes, with original audio when available and no artificial narration for chat studies; prompt does not select different quotes. Poll inspect_research_session view=job until succeeded; then read its output_id. Retry with the same key.

Read the effective workspace, role, available credit balance, payment method readiness, and incentive limits before a paid action. A payment method on file does not approve a Panel launch; use the current estimate and its MCP App for human approval.

Poll a queued study customization turn. On succeeded, inspect result and call get_study to return the persisted plan. On failed, inspect the saved study before retrying.

Read the customer-supplied panel entry link, respondent ID parameters, and outcome redirects for one owned BYOP study. This is distinct from individual BYOP invitations and paid User Intuition Panel recruitment.

Get one feasibility request and its status or estimate.

Get qualifying completions versus the current target, completed interview quality counts, and a rate-based expected finish when enough progress exists.

Get interview timing, stored format, current configured study language, and up to 100 spoken transcript segments bounded to about 20 KB of text. Use next_message_offset to fetch the next page. Each segment keeps its original turn ID and text_offset; system prompts and internal metadata are excluded.

Aggregate eligible non-test interview counts and billed units over a date range. The backend resolves the default app and a 30-day UTC range when omitted. The response includes the applied scope.

Read one BYOP invitation, including its current status, personal interview link, interview count, and reward state. Use it to verify the participant before changing their email, deleting the invitation, or sending a reward. Panel respondents are not returned by this tool.

Poll a participant batch. On success, list_participants for the study to retrieve the invitations.

Poll an asynchronous report request. A succeeded job provides result_url; then fetch the report with get_study_report.

Poll an editable PowerPoint export. A succeeded job returns a fresh private download URL valid for 15 minutes. The export expires after seven days.

Read a run and verified claims, its evidence, eligible timed highlights, exports, source coverage or session history. Follow next_cursor until exhausted. Evidence includes canonical transcript locators; audio eligibility requires a complete participant turn and owned recording. Authorization is checked on every read.

Poll a Hub export. A succeeded job returns a fresh private download_url with source access rechecked at download. Expired or revoked sources cannot be downloaded.

Get the full persisted study, including the plan, audience targeting, screeners, and concept assets managed by customize_study. Call before metadata updates and after customization to verify the result. After customization, return the complete persisted study_plan to the user in a readable form. Panel launch approval happens in the launch_panel MCP App, or explicit human chat confirmation when the client does not support Apps. Obtain explicit approval of the current plan before BYOP participant creation; retain study_id and updated_at so unchanged approval survives a resumed workflow. When provisioning_status is provisioned, include both returned links in the completion response. dashboard_url is the researcher management link. study_link is the live participant interview link, not a test or preview link; opening it starts the participant experience. The public API does not expose a separate researcher-preview URL.

Get a small report overview by default. Set view to one section (study_findings, participant_profiles, participant_responses, evidence_coverage, recommended_next_steps, or references), or full when all data is needed. included_sections distinguishes omitted sections from empty ones. Findings include structured prevalence and source quotes when available.

Get a webhook by ID without revealing its signing secret.

Read organization Hub sessions, one session, sources, chronological chat history, a durable job, or Studio outputs. Set session_id except when view=sessions; job needs job_id, output needs output_id. Follow next_cursor unchanged as after. Job success includes answer/references or output_id; partial remains explicit. Read the existing output before editing it.

Open the MCP App for human review of the saved study plan and complete current panel estimate. The first call only previews and does not spend. For App clients, a human must click Approve and launch in the App. For clients without MCP Apps, return the saved plan and quote in chat, ask the human to confirm, then call this tool again with the returned chat_approval_nonce and human_confirmed=true. Never infer confirmation from a prior study or quote. The estimate must match the study, country, inputs, balance and price at approval time.

List the account feasibility requests, newest first, in bounded pages.

List lightweight interview rows. Fetch an interview to get messages, recordings, and screener responses.

List organizations you can access. Users can belong to multiple organizations. Search by organization name, active owner name, or owner email; returns IDs and owner contacts. Ask the human which organization to use, then supply organization_id on subsequent tools.

List BYOP invitation records with each personal interview link, status and visible interview count. These are study participations, not deduplicated people; use list_interviews(participant_id) for the sessions. Panel and open-link respondents are anonymous interviews.

List lightweight studies, sorted by updated_at descending then ID. Filter before pagination by provisioning status, recruiting method, creation date, or report availability. interview_count is completed visible non-test interviews; use get_study for the full plan, screeners, and targeting.