When your AI agent moonlights as a UX researcher

What it does

This is a plugin-style “skill” that bolts user-research tooling onto AI agents like Claude, Codex, and Cursor. In standalone mode it drafts research plans, screening questionnaires, and interview guides, or processes existing transcripts to produce structured reports with themes and personas. To run live AI-moderated interviews or recruit real survey respondents, it relies on the external Cookiy AI platform.

The interesting bit

The project treats user research as just another agent capability rather than a separate workflow. It is primarily integration glue and configuration that teaches your coding assistant to ask structured questions and synthesize answers, though the heavy lifting for live participants happens on Cookiy’s servers.

Key highlights

• Generates research plans, screeners, and interview guides inside the agent chat
• Synthesizes raw transcripts into codebooks, personas, and prioritized findings via a five-phase pipeline
• Supports AI-moderated qualitative interviews and multi-language quantitative surveys through Cookiy AI
• Works across Claude (Desktop, Cowork, Code), Codex, Cursor, and OpenClaw
• README cites 900K+ creator views and 5,000+ comments from early users

Caveats

• End-to-end studies with real or synthetic participants require network egress to s-api.cookiy.ai; standalone mode is limited to planning and synthesis
• The repository appears to be primarily integration glue and skill configuration rather than the core research engine
• Social proof metrics (views, comments) are self-reported from creator content, not independent benchmarks

Verdict

Worth a look if you want your agent to handle the full research lifecycle from plan to report. Skip it if you need a fully self-hosted user-research platform—this one leans on Cookiy AI for the live participant layer.