Agent Skill and CLI · open source

Let any AI
talk to real people.

The open-source Cookiy Skill and CLI give Claude, Codex and Cursor a research team: real participants, AI interviews, surveys at scale and the report.

View on GitHub · ★ 1.5k Install

Your agent can write code, docs and plans. Now it can ask real people.

A user research skill for AI agents: Claude Code user research, a research CLI for Codex and Cursor, and a bridge to Cookiy AI for real participants. Qualitative interviews and quantitative surveys, from one prompt.

For AI agents

Qual and quant,
from one prompt.

The Cookiy Skill and CLI turn any question into a full research program inside Claude, Codex or Cursor: participants, interviews, surveys and the report. Open source, MIT licensed.

QualAI interviews with real people, from five to hundreds per study

QuantMulti-language surveys with screening, conditional logic and recruitment, at the scale you need

View on GitHub · ★ 1.5k

claude

> /plugin marketplace add cookiy-ai/user-research-skill
> /plugin install user-research@cookiy-ai
✓ installed cookiy user-research

> Interview 20 ICU nurses about shift handover,
  then survey 2,000 nurses in the US and UK to size it.

● Discussion guide + survey drafted .. done
● 20 ICU nurses verified, interviewed  20/20
● Survey live in 2 markets ...... 1,312/2,000
● Themes, tone and crosstabs ...... running
→ One report: the quotes and the numbers

Install

Two minutes to set up.

Claude Code

/plugin marketplace add cookiy-ai/user-research-skill
/plugin install user-research@cookiy-ai
/reload-plugins

Codex, Cursor, OpenClaw and others

npx cookiy-ai

Claude Desktop and Cowork: add the marketplace cookiy-ai/user-research-skill in Customize, or upload the skill ZIP. Full steps on GitHub.

Then type /user-research-cookiy, or just describe your research goal.

What to ask it

Four things people ask first.

“I want to understand why users churn after onboarding.”

Designs a research plan, screener and interview guide targeting early drop-off.

“Here are 20 interview transcripts, give me a report.”

Runs a five-phase synthesis and returns themes and personas.

“Run a study on how developers choose CI/CD tools.”

Creates the study in Cookiy AI, runs AI-moderated interviews and delivers the report.

“Survey 200 users about feature satisfaction.”

Designs the survey, recruits respondents and collects the results.

Keep exploring

Related: Surveys, AI user interviews and Platform overview.

SurveysQuantitative surveys at scale, fielded to the right people.Learn more →

AI user interviewsIn-depth interviews with real people, a hundred a day if you need them.Learn more →

Platform overviewEvery part of a study, from recruiting to the report, in one place.Learn more →

About the skill and CLI.

What is the Cookiy user research skill?

An open-source, MIT-licensed skill for AI agents. It teaches Claude, Codex, Cursor and other agents to plan research, write guides and screeners, synthesise transcripts, and, connected to Cookiy AI, run real studies with real participants.

Which agents does it work with?

Claude Code, Claude Desktop and Cowork, Codex, Cursor, OpenClaw and other agents that support skills. Install instructions for each are on GitHub.

Can the agent run real interviews and surveys?

Yes, through Cookiy AI: it creates the study, recruits participants, runs AI-moderated interviews or fields a survey, and returns the report.

Does it work without a Cookiy account?

The general research features (plans, guides, synthesis of your own transcripts) work on their own. Running studies with participants needs a Cookiy AI account, which is free to start.

Can it run large surveys?

Yes. Ask your agent for a survey of the size you need and Cookiy designs it, recruits respondents and collects results.

Is it really open source?

Yes, MIT licensed, with 1,500+ stars on GitHub. Contributions are welcome.