Developers · Open source
Two open‑source skills.
One each way.
One lets your agent reach real people. The other lets people sell the data they already have. Both run in Claude Code, Codex, Cursor and OpenClaw.
Agents can ask real people. People can sell what they already made.
Both skills are small, public and installable from the terminal. They are the open ends of the same supply chain we use to source human data for frontier labs: one brings real people to your agent, the other brings contributors to buyers.
- Works with. Claude Code, Codex, Cursor and OpenClaw.
- Open source. Code, install steps and docs live on GitHub.
Open source
The two skills.
Agent → real people
User Research Skill
Ask a research question from your agent. The skill recruits real people, runs AI-moderated interviews or surveys, and returns structured findings to the agent.
- Recruit by profile, market and behavior
- AI interviews, synthetic users and quantitative surveys
- Synthesis and structured answers back to the agent
People → data buyers
Sell Sessions Skill
Sell the Claude Code and Codex sessions you already ran. A local uploader lets you pick every session, scrubs secrets and PII on your machine, and pays you through Cookiy Earn on every sale.
- Runs locally, MIT licensed
- You choose which sessions leave your machine
- Paid on every sale
Behind the skills
Where the skills connect.
Questions teams ask.
Which agents do the skills work with?
The User Research Skill works with Claude Code, Codex, Cursor and OpenClaw. The Sell Sessions Skill reads the Claude Code and Codex sessions already on your machine.
Is there a general-purpose API?
Not yet. The two skills are the public entry points today. An agent that spots a data gap, requests real human evidence and feeds it back into evaluation is a research direction, not a documented API.
I need more than the skills.
Broader sourcing, licensing or annotation starts with a brief. Start one here.
Need more than the skills?
Start with a brief.
Video on this site is licensed stock footage of real people, used to illustrate the kinds of data we collect.