How the Hub works
The mechanics, stated plainly — no fine print.
What it is
A community library of prompts that actually produced something. Every media card pairs a real result with the exact words that made it. Text posts ride the same feed, so you can participate without generating anything.
Publishing
Two ways in. From the studio: one tap on any result — the card mints with the prompt and model attached. Or a text post from /hub/submit. Both go through automated review before they appear.
Ranking
Hot is engagement over age decay — views, likes, comments, shares, and above all uses: how many times the prompt was actually run in the studio. A prompt people use outranks a prompt people applaud. Latest is plain chronological.
What creators get
A public profile page — your handle shows your cards, total uses, likes, and followers, and a Follow button lets people subscribe to you. Top creators rank on the hub front page, weighted by real usage. Revenue share isn't live yet — when it arrives, it will be calculated from paid prompt runs, not clicks or follower counts.
Moderation
Automated, fail-closed — text is screened and media classified at publish time; anything unsafe or ambiguous is rejected, never queued for a human. Cards rated for general audiences are also the only ones exposed to AI agents through MCP. Members can report; five unique reports unpublish a card automatically.
For agents
The same library is readable by AI agents over MCP at livepairai.com/mcp — search, trending, and per-card fetch tools, listed in the official MCP registry. Setup for Claude Code, Cursor, VS Code, Gemini CLI, and Zed is in the MCP docs.
The rules
No explicit sexual content, no spam, no prompts built to attack other users. Racy-but-safe work stays on the site; it just never travels to agents. If a card is yours, you can unpublish it anytime.