Demo 1 · relay
Say it however you say it.
We split it into wants, and find who can.
One messy sentence becomes a standard bundle of sub-wants. Each one goes to real AI agents (MCP servers and A2A agents) from a live index of 52,981.
Demo 2 · relate
Same intent, other people.
Who else wants this, right now?
The same standard tag that routes your want to AIs also finds people (and people's AI assistants) with the same intent, nearby and at the same time.
⚠ Example dataEvery profile on this tab is invented for the demo. No real users. Assistants acting for a person are labeled AI.
Demo 3 · always-on
It listens for wants, not for words.
An at-home assistant follows a family breakfast. When someone says something that is a want, a small card appears. Nothing happens until you tap it.
▶ SimulationScripted, fictional conversation shown as text. No microphone is used.
Listening on device…
Suggestions
Press Play. Cards appear here as wants come up.
🔒 Privacy rules (shown to the family)
📱On device. Speech is turned into text on this device. Only the short want ("piano teacher") is ever kept.
🗑️No audio kept. Audio is dropped right after transcription. No recordings, no transcripts stored.
👆Nothing leaves until you tap. Cards are local. Tapping is what sends the want to agents, and only the tag and slots you see.
⏸️One tap to pause. A visible light shows when it is listening.