A qualitative research API lets your product commission customer interviews and retrieve evidence without building its own recruitment and moderation systems. Your application owns the interface and the customer relationship. User Intuition supplies study planning, recruitment, interviews, and research outputs.
Start with the public TypeScript examples and API reference. Fixture mode requires no account and makes no network calls. It is a demonstration with fictional data; check the repository’s compatibility notes before using live results or search adapters.
What does the integration look like?
- Connect a server-side account. Create an organization-scoped API key and store it in your secret manager. Keep credentials out of browser code, logs, and source control.
- Create and customize a study. Create the study metadata, then send the research brief through Customize Plan. Handle follow-up questions and read the saved plan before proceeding. A study title alone does not define the interview.
- Review the plan and audience. Show the user what will be asked, who should participate, and how participants will be recruited. Your application should make these choices understandable before any invitations or spending.
- Estimate recruitment. For a panel study, request the estimate for the proposed audience and sample. Platform interview charges and recruitment are separate. Feasibility depends on the audience, country, and study settings.
- Launch approved fieldwork. Preserve the user’s authorization and selected settings. An explicit example launch flag demonstrates the interaction; it is not a replacement for your product’s approval system.
- Monitor progress. Read study and interview state. Do not treat a configured sample target as the number of completed or analyzed interviews.
- Retrieve and present results. Read the four report sections and follow source references to interviews. Keep coverage and freshness visible alongside the finding.
The example exposes these as separate actions so developers can inspect each step. It does not silently recruit people when someone installs dependencies or runs the default demonstration.
What should your product display?
Study Findings summarize the research. Participant Responses describe individual answers. Participant Profiles describe the sample. Recommended Next Steps suggest actions or further research. These sections serve different purposes and should retain their labels in your interface.
A useful results screen pairs each finding with its source references and the relevant study context: audience, research dates, and analyzed sample where supplied. An answer summary is not a verbatim quote. To present a quotation, retrieve the interview and use the referenced message or exact passage when that granularity is available. An interview-level reference alone does not identify an exact sentence.
Your application determines how to combine findings with product analytics, account context, and other evidence. The API does not require User Intuition to own that broader reasoning layer. See the results example for a fictional reference-resolution walkthrough.
How can the product reuse existing research?
Search relevant findings and participant responses across authorized studies before proposing additional fieldwork. Keyword matching helps locate exact terms; semantic matching helps locate related ideas. Inspect each result’s study context and retrieve its report or interview before relying on it.
Search coverage matters. A search over findings and response summaries is different from searching every transcript message. Preserve the returned coverage and filters rather than presenting a result as an exhaustive review of the entire research library. No matches, incomplete indexing, and a failed request are different states in your interface.
The search example shows result handling and source retrieval guidance. Your product decides whether the retrieved evidence answers the question or whether another study would be useful.
What changes when you serve multiple customers?
Define authorization and data ownership before exposing research in a multi-customer product. Organization-scoped keys are a platform boundary; your own application must also prevent customers from requesting each other’s study IDs or viewing cached responses belonging to another account.
Confirm supported account setup and commercial terms with User Intuition. Do not design around an assumed public organization-provisioning API. Use separate credentials and access boundaries where your agreed setup requires them, and preserve those boundaries in background jobs, search, exports, and support tooling.
Researchers can use the dashboard while engineers integrate documented operations. Decide which actions your own interface supports and which require dashboard access. API and dashboard capabilities may differ; the operation reference is the authority for your implementation.
How should the integration handle failures and costs?
Handle errors explicitly. A timeout during a write does not establish that the operation failed. Read persisted state before repeating study creation, invitations, or a panel launch. Do not assume every write is idempotent or add automatic retries without checking the operation contract.
A stale report may exclude recent interviews. Show freshness and sample coverage; request report generation separately when appropriate. An unavailable report should not become an empty “no findings” screen.
For current pricing, separate platform usage from participant recruitment. Starter voice interviews cost $30 each. Recruitment depends on the audience and study; inspect the actual estimate. Reseller and volume arrangements require agreed commercial terms.
Where should developers start?
Run the examples, inspect the responses, and choose the operations your product needs. The research infrastructure overview explains the product-builder use case. The agent research overview covers an agent conducting research within a broader task. Both rely on the same principle: retrieve inspectable research evidence, then let the calling application decide how to use it.