Viewers submit clues privately. AI drafts three tiers grounded only in what they sent, and each tier lists the clues behind it. You approve, edit, skip or regenerate every tier before it reveals. Hidden tiers stay on the server.
AI can still be wrong or too revealing. That's why every tier needs your approval first.
Flags copy-paste waves, floods, near-duplicates and repeated promo links, and says why each one was flagged. You can dismiss a flag, confirm it, or delete the message through Twitch's moderation API.
Never bans anyone automatically. Works alongside Twitch AutoMod, not instead of it.
Private summaries of recent chat: main topics, questions people keep asking, useful suggestions and reactions. Every item links back to the messages behind it, and spam is left out.
Built only from messages Breadcrumb actually received. Counts are computed, not guessed.