WiFi meeting audio endpoint
The room microphone for AI agents.
Hummic records 3-5m meeting-room audio, uploads the file to your HTTP server, and can stream live audio to your agent runtime.
- 3-5m
- Room pickup target
- HTTP
- File upload first
- Agent
- Live stream ready
Two operating modes
Record files when meetings matter. Stream audio when agents need ears.
Small screen, real control
Pair by QR. Read status. Get short agent replies.
Hummic's display is intentionally small: it shows pairing codes, WiFi state, upload state, endpoint names, and concise agent responses. It is not a tablet, and it is not a locked-in assistant.
- QR pairing for device setup and runtime binding
- Recording, uploading, synced, and failed states
- Short text replies from your agent runtime
- Local cache with retry when the network is unreliable
Workflow first
Hummic does the capture. Your stack does the intelligence.
Meeting-table microphone design for real room pickup, not close-talk dictation.
Post complete files to an endpoint or stream audio to a live agent session.
Transcribe, translate, summarize, email, update CRM, or write to a knowledge base.
Industrial design directions
Quiet hardware for the center of a table.
The design language is closer to a professional network device than a smart speaker: low profile, physical control, visible microphone geometry, USB-C power, and one restrained blue status line.
Developer surface
Plain HTTP first. Agent streaming next.
File upload
POST /upload
Content-Type: multipart/form-data
file=@meeting_20260528_153000.m4a
device_id=hummic-room-01
started_at=2026-05-28T15:30:00+08:00
ended_at=2026-05-28T16:12:00+08:00
duration_sec=2520
format=m4a
sample_rate=16000
Agent channel
WS /agent/live
client -> audio_chunk
client -> button_event
client -> status
agent -> text_reply
agent -> session_state
agent -> display_hint
Positioning
Built for builders, not locked-in apps.
No mandatory Hummic cloud.
No bundled meeting-note workflow.
No forced AI provider.
Your audio. Your endpoint. Your agents.