Built on Qloo’s taste data
Be ready before the doorbell rings.
A child arrives tonight and you know almost nothing. Share a few favorites. An agent builds a small, kind plan of things to watch, play and eat, and shows you everything it held back and why.
- No names or case details
- Nothing is stored
- Age, triggers and sensitive themes checked in code
Your plan appears here
Pick an example on the left, or fill in what you know. It takes about fifteen seconds.
How it works
- Qloo knows taste. Each favorite is looked up in Qloo. Its recommendations, tags and explainability scores supply every title, artist and restaurant. Nothing is invented.
- An agent plans. A Gemini model chooses which tools to call and words the openers and the evening. It sees only what you typed after the privacy screen.
- Code has the last word. Content ratings for the child’s age, trigger words, themes like separation or adoption, explicit music and adult venues are enforced in code. The agent cannot override them, and every item held back is listed with its reason.
For agents
The same planner and safety layer is a Model Context Protocol server over Streamable HTTP. Any MCP client can plan a first night or check one title.
Endpoint
/mcpTools
plan_first_night, suggest, nearby_food, check_item