Rival MCP lets you connect an AI agent you already use, directly to your Rival research data. Once connected, you can ask questions about your studies, dig into what respondents said, and pull data for analysis in plain language, without logging into Rival, exporting files by hand, or building a custom integration.
MCP stands for Model Context Protocol, an open standard for connecting AI agents to external tools and data sources. Rival MCP is our implementation of that standard, scoped to your account's research data.
How access works
Rival MCP authenticates you as a Rival user: your agent doesn't get its own separate access, it gets exactly what you can already see in Rival. To connect at all, your Rival account needs the Research Domain Admin role on the research domains you want to work with; without it, the connection will succeed but every request will be refused.
What you can do with it today
Rival MCP is currently read-only, meaning your agent can look up and retrieve data, but it can't create, edit, launch, or delete anything in Rival. Within that scope, here's what's available:
Find and understand your studies
List the research domains and studies you have access to
Pull a study's profile (sample size, field dates, question types) before deciding whether to dig in further
Get a study's full question list, or its complete script with display logic, piping, and masking rules
Find studies by matching question text or answer-choice wording, without knowing which study to look in first
Read and search responses
Pull every response to a specific question, or the full response set for a study
Search across what respondents actually said, across studies, by topic or keyword
Export larger response sets as a file when there's too much to work with inline
Look up and segment participants
Look up an individual participant's profile attributes
Search for participants matching a set of profile conditions (e.g. region, age range, plan type) and export the segment
PII-flagged attributes and responses are automatically redacted in whatever comes back; MCP can still filter and search on them but your AI agent never sees the underlying value
See how a study was fielded
Pull distribution data: how a study was sent out, its entry points, and delivery timing
Check engagement counts and individual delivery status (completed, incomplete, failed, and so on)
Export delivery and engagement data for larger studies
Pull participant-uploaded media
Export the images and videos participants uploaded as open-ended answers
Example use cases
"Which of our studies asked about pricing sensitivity in the last year?" Search across studies by topic instead of hunting through a list.
"Summarize what respondents said about the onboarding experience in [study]" Pull and synthesize open-ended responses without exporting to a spreadsheet first.
"Pull everyone in the West region under 40 who completed [study], and export it" Segment participants by profile attributes and get a working file.
"How did the [study] wave perform: completion rate, drop-off, any failed deliveries?" Check fielding health without opening the platform.
"Compare sentiment on the support experience question across our last three NPS waves" Cross-study analysis using sentiment scoring already computed on open-ends.
"Pull the images customers uploaded for the packaging feedback question" Grab participant-submitted media directly.
What's not supported yet
Write access, meaning using your AI agent to create or launch a study, edit questions, or modify response data, isn't available yet. Rival MCP is intentionally read-only in this phase; write access is something we'll be working on in an upcoming phase. We'll update this article when that changes.