← Reference Deep-Dives Reference Deep-Dive · Updated · 8 min read

Cross-Study Qualitative Research: Query Past Research via MCP

By Kevin, Founder & CEO

Research compounds when each new decision starts from everything your team has already learned. An AI agent connected to User Intuition makes that practical: ask a question across your past studies, get an answer cited to participant quotes, and commission new interviews only for what the library cannot answer. This guide is the full how-to: choosing what a session covers, writing questions the library can answer, reading the citations, running a full analysis when a count matters, and turning the answer into a deliverable or the next brief.

Why does research compound instead of resetting?

Most studies are read once, presented once, and filed. The next team with a related question starts from zero, briefs a new study, and pays again for evidence the organization already holds.

Cross-study research reverses that. Before commissioning new interviews, an agent checks what earlier studies already say about the audience, product, or decision. Studies on packaging, pricing, brand perception, and onboarding stay useful long after fielding ends, because each interview is kept with its transcript, recording, and quotes. A finding from last spring’s shopper study can answer part of this quarter’s question, and the agent can show you exactly which participants said it. The gain is more than budget. A question answered from the library comes back in the same conversation, before a new study would have finished recruiting, and it arrives with evidence the team has already read and trusted. When the library answers only part of the question, the remainder becomes a tighter brief: a smaller study aimed at what is genuinely unknown, instead of a broad study that asks again what the organization already learned.

How does an agent ask a question across your research library?

Connect it in Claude, ChatGPT, Codex or Cursor in a few minutes (setup), then ask in plain language, for example: “What do shoppers say about comparing prices before they choose a store?”

The agent opens a research session, adds the studies you choose or every study in your account, and asks your question. The answer comes back as written prose with citation labels, and each label opens the participant quote, the study, and the interview it came from. In Claude, which supports MCP Apps, each cited quote opens as a card with the transcript and the original recording inside the conversation.

For a quick question about named studies, two lighter tools skip the session. search_research returns the matching records grouped by study, and can be narrowed to findings, participant responses, profiles, study plans or recommended next steps, and to a research date range. answer_research returns generated statements, each with numbered citations to those records.

For deeper work, the session holds a running conversation. Follow-ups such as “which of these came from first-time buyers?” build on earlier answers, and citations are preserved through every turn.

Which studies and sources should a session include?

The scope of a session decides what the answer can draw on, so set it on purpose.

  • Every study, or a chosen set. Start with every study in your account when the question is broad. Name the studies when it concerns one product, market or period, because a narrower session gives a sharper answer.
  • Studies added later. Add a study to an open session at any point, and resync one when it gains new interviews, so the answer reflects the current evidence.
  • Your own documents. Attach earlier agency reports, survey exports or support-ticket summaries as TXT, Markdown, CSV, JSON, PDF or DOCX files, and the answer draws on them beside the interviews.
  • Removing a source. Take a study out of the session without deleting it from your account, or clear the session’s sources and start again.
  • Real interviews by default. Synthetic respondents are also available for a directional first read, and their answers stay out of a session unless you ask to include them.

One habit keeps the library useful: open one session per decision and name it for that decision. Follow-ups stay in context, and anyone on the team can reopen the thread and its history later.

How do you write a question the library can answer?

A cross-study question works best when it carries what a researcher would put in a brief: who, what behavior, which period, and what kind of answer.

  • Name the audience and the behavior. “Among first-time buyers, what made them pick our brand over the store brand?”
  • Bound the period. “In studies since the spring relaunch, what do shoppers say about the new pack?”
  • Ask for counts by person. “How many participants raised price as a reason to switch, and in which studies?”
  • Ask for disagreement. “Where do our studies contradict each other about the refill format?”
  • Ask for the gap. “Which parts of this question do our studies leave unanswered?”

Each form keeps the answer tied to evidence rather than to a summary of summaries. The last one is where the next brief starts.

What does a cited answer contain?

A cross-study answer is built to be checked, not taken on faith.

  • Plain-language answer. A direct response to the question, written from the evidence in the session.
  • Citation labels. Each claim carries labels that point to the transcript turns behind it.
  • The quote in context. Opening a citation shows the exact quotation, the surrounding transcript, the study name, and the original recording where one exists.
  • Honest counting. Citations are grouped by source interview, so several quotes from one participant count as one person.

That structure lets a reader trace any sentence that will drive a decision back to the person who said it.

When does a question need a full analysis?

A chat answer retrieves the most relevant evidence quickly. Some decisions need more than that: every interview read, or an exact count across studies. For those, the agent can run a durable analysis over the session’s sources.

  • Exhaustive reads every eligible interview in the session, for questions where missing one participant could change the answer.
  • Aggregate computes exact counts, such as how many participants across four packaging studies said the pack shaped their choice.
  • Quick retrieves verified evidence for a fast first read before you commit to a longer run.

You can hand the analysis a codebook of the themes you care about, group the results, for example by study, and ask for the result as an answer, a report, a table or a presentation. A finished run can also produce clips and highlight reels from the moments it cites.

What does the underlying evidence look like?

Answers are only as good as the studies behind them. Here is example output from a User Intuition study of 43 Walmart shoppers, each reconstructing one recent trip in a voice interview. The headline finding reads: “This is not loyalty. It is tolerance, and it is priced.”

36 of 43 shoppers will shop there again, and W07 described how that kind of choice gets made:

“So I compared prices between Walmart, Target, Amazon, eBay, and the best reasonable price and closest distance was Walmart.”

Findings like these, each with a count and the verbatim quotes behind it, are what later cross-study answers cite. A question about what decides the store across several retail studies can cite this finding and open W07’s quote in context. Explore the Walmart study to read the findings and listen to sample interviews.

What does a cross-study workflow look like?

  1. Choose the scope: every study, or the ones the decision depends on, plus any documents.
  2. Ask the question in plain language and read the cited answer.
  3. Open the source interviews behind any claim that will drive a decision.
  4. Narrow with follow-ups by audience, market or period, or compare two studies, such as before and after a packaging change.
  5. Run a full analysis when the decision needs every interview read or an exact count.
  6. Turn the answer into a deliverable, or brief a new study for the gap.

Which questions suit cross-study research?

Has the reason shoppers switch brands changed? Compare studies from different quarters and read the quotes behind each shift.

Does a claim that won in one market hold in another? Studies run in 80+ languages, so the agent can set findings from different markets side by side.

Which packaging and messaging themes recur? Pull concept-test findings from several rounds and see how often each theme appears, as n of N.

Did an onboarding change fix the earlier problem? Compare interviews before and after the change.

What should the next study investigate? Gather unanswered questions and contradictions, then let the agent draft the next brief from them.

How do you turn an answer into a deliverable?

The research session that answered the question can produce the output: a written report, a native PowerPoint deck or a highlights video, with citations carried in so the deck shows where each point came from. Upload a report template in Markdown, TXT or DOCX, or a 16:9 PowerPoint template, and the output reuses its structure and theme. The MCP research API reference covers every deliverable and download format.

What happens when the library cannot answer?

When the evidence falls short, the agent drafts the next brief from the gap: the open questions, the audience they concern, and the contradictions worth settling. That brief starts a new study, and the MCP research API walkthrough covers the run from brief to cited report. Once the new study finishes, add it to the same session, and the next answer draws on old and new evidence together.

How does the Customer Intelligence Hub fit?

The Customer Intelligence Hub is the library an agent queries, and the dashboard lets your team explore the same studies visually. See the research API and MCP server for AI agents for everything else an agent can do with it.

The library is worth querying because of how it was collected. User Intuition’s AI moderator runs 20–30 minute interviews that ladder 5–7 levels deep, so the library holds reasons rather than one-line answers. Every panelist call is reviewed by the User Intuition team, panelists pass a video identity check, and only interviews that pass quality scoring are billed; see how quality works.

Get started

Connect your agent, then ask a first question across two or more studies. When you commission the next study, recruit from a vetted 4M+ consumer panel (B2B audiences recruited to order per study), or invite your own participants. Voice interviews cost $30 each, and signup includes 3 free interviews with your own participants; panel recruiting is billed separately.

Note from the User Intuition Team

User Intuition provides AI-moderated qualitative research for agencies, consulting firms, and research teams. Keep your methodology and discussion guide, bring your own sample or use our 4M participant panel, and review recordings, transcripts, and evidence-linked findings. Your researchers connect the evidence to the client decision and prepare the final recommendations.

Inspect complete sample calls and a readout, then test your own brief. Starter voice interviews cost $30 with your sample or $60 with standard panel recruitment, with no monthly fee. Specialty audiences are quoted separately; incentives you arrange for your own sample are additional. See pricing or try 3 free voice interviews with your own participants.

Frequently Asked Questions

Cross-study research compares evidence across several studies to reveal recurring themes, shifts over time, and differences between audiences or markets. An agent asks one question of your research library and gets an answer cited to the quotes behind it.

Ask your agent a question in plain language. It opens a research session over the studies you choose, or every study in your account, and returns an answer cited to participant quotes, each linked to its source interview.

The Customer Intelligence Hub is the research library where your studies, findings, and quotes are kept and explored. MCP and the API give agents the same library to query, and the dashboard lets your team explore it visually.

A cross-study query retrieves evidence from your own interviews and answers from it, so each statement is cited to a participant quote you can open. A general model answer draws on training data instead of your research.

Yes. After the agent retrieves the answer, ask for a written report, a PowerPoint deck, or a highlights video from the same research session, with citations carried into the output.

A research session can include every study in your account or only the studies you select. You can also attach documents such as earlier agency reports or survey exports as TXT, Markdown, CSV, JSON, PDF or DOCX files, and the answer draws on them beside the interviews.

Yes. Once connected, an agent can search and cite at any time. When a gap in the library calls for new research, the agent shows you the saved plan and a cost estimate before anything is spent; in Claude you approve with one click, elsewhere you confirm in the conversation.

A direct query answers one question about the studies you name. A research session keeps a running conversation over the studies you add, so follow-ups build on earlier answers and keep their citations.
Get Started

Put This Research Into Action

Run your first 3 AI-moderated customer interviews free with your own participants — no sales call.

Self-serve

Launch your first study in minutes. Results in 24 hours.

See it First

Explore a real study output — no sales call needed.

No contract · No retainers · Panel results in 24 hours