Cross-Study Customer Research: Querying Past Research via MCP

Research that compounds instead of resets

Most research teams work with a recurring problem: each study starts from scratch. The findings from last quarter’s churn interviews don’t connect to the findings from this quarter’s positioning test. The themes a team found in early-stage discovery don’t surface when a PM is writing a spec six months later. Research accumulates in slide decks and folders rather than in a queryable knowledge store.

The consequence is redundant fieldwork. Teams re-interview similar segments to answer questions their past research already touched, not because they want to duplicate work, but because there is no efficient path to what was already learned.

Cross-study customer research solves this at the architecture level. When every completed interview feeds a shared knowledge store, and an AI agent can query across the full corpus, the value of each new study compounds with the existing library. Research becomes an organizational asset that grows rather than a series of disconnected point-in-time outputs.

For teams building AI-native research workflows through the User Intuition agentic research platform, the Intelligence Hub is the infrastructure layer that makes this model work.

Why most research tools don’t support cross-study queries

Research platforms are typically designed around the study as the unit of organization. Navigation is per-study. Exports are per-study. Findings are tagged to studies. This design is sensible for teams running occasional, standalone projects — it matches the workflow of planning a study, running it, and reporting on it.

The design breaks down when organizations run research at higher frequency, as AI-moderated platforms make possible. When a team runs 10-20 studies per year instead of 3-5, the per-study silo creates a library problem: the older studies contain relevant signal, but there is no practical way to query across them. The institutional memory stays locked in documents no one reads after the original stakeholders move on.

Several structural gaps prevent most platforms from solving this:

No shared corpus. Studies are stored independently, not as parts of a unified index. A search interface might exist for a single study’s transcripts but not across all studies in an account.

No grounded retrieval. General-purpose LLM interfaces layered onto research tools can answer questions about your research, but they generate responses from their training data rather than retrieving from your actual transcripts. The answers sound authoritative but are not sourced from your specific studies.

No compounding architecture. Each study produces a report. The report sits in a folder. There is no layer that continuously indexes incoming interview data and makes it queryable as the corpus grows.

How MCP enables cross-study queries

The Model Context Protocol gives AI agents a standardized interface to call tools on external systems. For cross-study research, this means an agent can call Intelligence Hub tools the same way it calls any other capability — with a typed manifest, parameter documentation, and a clear response contract.

The User Intuition MCP server exposes 18 Intelligence Hub tools, organized into three functional areas:

Grounded retrieval. The query_intelligence tool accepts a natural-language question and returns an answer grounded in the actual content of your Intelligence Hub corpus — verbatim excerpts, source attribution, and participant metadata. This is not an LLM generating a plausible answer; it is retrieval from real interview data.

Session management. The create_session, list_sessions, get_session, update_session, and delete_session tools manage conversational context for multi-turn cross-study analysis. An agent can open a session, ask a series of connected questions about past research, and maintain context across turns.

Output generation. The generate_report and generate_powerpoint tools produce structured deliverables from cross-study queries. Both run as background jobs and are retrieved when complete. The create_studio_output, list_studio_outputs, get_studio_output, and update_studio_output tools manage saved analysis outputs within the Hub.

The full Intelligence Hub group also includes document management tools (list_documents, delete_document, get_file_search_store) and chat history tools (list_chat_history) for audit and debugging purposes.

4 example queries an agent might ask

These represent the kinds of cross-study questions that become practical when an agent has Intelligence Hub access via MCP.

“What is our cumulative churn-driver picture?”

Query the Intelligence Hub for all interview findings related to 
cancellation and churn across all studies in the workspace. 
Summarize the top five drivers with supporting participant quotes 
and indicate which studies each theme appeared in.

A single churn study gives you a snapshot. A cross-study query gives you the pattern — which drivers are consistent, which are study-specific, and which have grown or declined in frequency over time.

“What messaging themes have resonated this year?”

Search past studies for participant responses to messaging and 
positioning questions. Identify which value propositions generated 
positive reactions and which generated skepticism or confusion. 
Group by time period to show how the pattern evolved across the year.

This kind of longitudinal analysis would require exporting every study to a spreadsheet and coding themes manually. Via MCP, it is a single tool call.

“Find every quote about pricing from the past six months”

Retrieve all participant quotes referencing price, cost, budget, 
or value from interviews completed in the last 180 days. Include 
participant segment metadata (role, company size) with each quote.

Verbatim quote retrieval with metadata is one of the highest-value Intelligence Hub use cases — the output is ready to use in a pricing deck, sales playbook, or copy brief without further synthesis.

“Compare our positioning evolution across launches”

For each major product launch in the past 18 months, identify 
the primary positioning messages tested in research and show 
how participant reactions to those messages changed from 
launch to launch. Highlight any consistent gaps between our 
messaging intent and customer understanding.

This query spans multiple studies and multiple time periods. Without cross-study infrastructure, answering it would require a research analyst to manually review every relevant study. With the Intelligence Hub, it is a structured retrieval query.

How does User Intuition handle cross-study customer research?

The Intelligence Hub is User Intuition’s answer to the silo problem — built directly into the platform rather than bolted on as a reporting add-on.

Every interview completed in a User Intuition study is automatically indexed into the workspace’s Intelligence Hub. There is no manual export, no integration setup, no separate tool to open. As soon as an interview completes, its content is available for cross-study queries.

Three capabilities define how the Hub handles accumulated research:

Grounded answers, not generated summaries. The query_intelligence tool performs retrieval-augmented generation against your specific corpus. When an agent asks about churn drivers, the answer is drawn from the actual words participants used in your interviews — not from a language model’s statistical approximation of what churn-related research typically finds. Source attribution is included so every finding can be traced to its origin study.

Session-based exploration for complex queries. Multi-turn sessions let an agent work through a research question incrementally — starting broad, narrowing to a specific segment or time period, drilling into a particular theme, and requesting supporting evidence — all in connected context. This matches how analysts actually think through research questions rather than forcing every question into a single-prompt format.

Report and presentation generation at scale. The generate_report and generate_powerpoint tools convert cross-study findings into structured deliverables without manual synthesis. For teams that need to share accumulated research with stakeholders on a regular cadence, this transforms the Intelligence Hub from a query tool into a publishing system.

The full Intelligence Hub is accessible via the User Intuition agentic research platform, with complete MCP tool documentation at docs.userintuition.ai/mcp-server/overview.

Get started with cross-study queries

To begin querying across your accumulated research:

1. Connect the MCP server. For Claude Desktop or Cursor, add the User Intuition server block to your MCP config and set your ui_sk_ API key. See the full setup guide.

2. Run your first cross-study query. Ask your AI agent: “What does our past research say about [topic]?” The agent will invoke query_intelligence against your Intelligence Hub and return grounded findings.

3. Start a session for complex analysis. For multi-turn exploration, ask the agent to open an Intelligence Hub session. This maintains context across a series of connected research questions.

4. Generate a report. For findings you want to share, ask the agent to call generate_report on a topic. The system produces a structured report from the full corpus, available in minutes.

As your study library grows, every new interview increases the value of cross-study queries. The research compounds — which is the point.