Great Question — MCP vs UI capability map

What the agent does today, where humans intentionally stay in the loop, and what's on the roadmap toward end-to-end agentic participant management.

MCP tools

~115

Repository read

Strong

Study authoring

Deep

Human checkpoints

By design

Agent does itHuman in the loop — todayOn the roadmapRResearcherPMProgram manager

Agent-led, human-supervised

The agent does the heavy lifting; humans stay in the driver's seat at the moments that touch a real participant or warrant in-app review — with end-to-end agentic participant management on the roadmap.

Repository — gets data to the agent

Sessions, transcripts, highlights, reels, insights, screener responses — the full evidence chain is readable.

Studies — drafts & configures

Interview, survey, unmoderated. Screeners with qualification + skip logic, incentives, consent, calendar, email copy.

Candidates — sources & triages

Search, create, update, segment, add to a study, send screener invitations, score responses.

Human owns the participant-facing moments — today

Launching a study, approving who advances, and direct outreach stay with a person, heading toward agent control.

The story: a deliberate division of labor that's actively expanding. The agent removes the busywork now; the human keeps judgment over anything a candidate sees — and the roadmap moves those handoffs into the agent's hands.