Zoom RTMS Meeting Assistant

Headless capture service for Zoom meetings using Real-Time Media Streams (RTMS). Receives webhook events, connects to RTMS WebSockets, records all media, and runs AI analysis via OpenClaw.

Webhook Events Handled

This skill processes two Zoom webhook events:

• meeting.rtms_started — Zoom sends this when RTMS is activated for a meeting. Contains server_urls, rtms_stream_id, and meeting_uuid needed to connect to the RTMS WebSocket.
• meeting.rtms_stopped — Zoom sends this when RTMS ends (meeting ended or RTMS disabled). Triggers cleanup: closes WebSocket connections, generates screenshare PDF, sends summary notification.

Webhook Dependency

This skill needs a public webhook endpoint to receive these events from Zoom.

Preferred: Use the ngrok-unofficial-webhook-skill (skills/ngrok-unofficial-webhook-skill). It auto-discovers this skill via webhookEvents in skill.json, notifies the user, and offers to route events here.

Other webhook solutions (e.g. custom servers, cloud functions) will work but require additional integration to forward payloads to this service.

Prerequisites

cd skills/zoom-meeting-assistance-rtms-unofficial-community
npm install

Requires ffmpeg for post-meeting media conversion.

Environment Variables

Set these in the skill's .env file:

Required:

• ZOOM_SECRET_TOKEN — Zoom webhook secret token
• ZOOM_CLIENT_ID — Zoom app Client ID
• ZOOM_CLIENT_SECRET — Zoom app Client Secret

Optional:

• PORT — Server port (default: 3000)
• AI_PROCESSING_INTERVAL_MS — AI analysis frequency in ms (default: 30000)
• AI_FUNCTION_STAGGER_MS — Delay between AI calls in ms (default: 5000)
• AUDIO_DATA_OPT — 1 = mixed stream, 2 = multi-stream (default: 2)
• OPENCLAW_NOTIFY_CHANNEL — Notification channel (default: whatsapp)
• OPENCLAW_NOTIFY_TARGET — Phone number / target for notifications

Starting the Service

cd skills/zoom-meeting-assistance-rtms-unofficial-community
node index.js

This starts an Express server listening for Zoom webhook events on PORT.

⚠️ Important: Before forwarding webhooks to this service, always check if it's running:

# Check if service is listening on port 3000
lsof -i :3000

If nothing is returned, start the service first before forwarding any webhook events.

Typical flow:

1. Start the server as a background process
2. Zoom sends meeting.rtms_started webhook → service connects to RTMS WebSocket
3. Media streams in real-time: audio, video, transcript, screenshare, chat
4. AI processing runs periodically (dialog suggestions, sentiment, summary)
5. meeting.rtms_stopped → service closes connections, generates screenshare PDF

Recorded Data

All recordings are stored organized by date:

skills/zoom-meeting-assistance-rtms-unofficial-community/recordings/YYYY/MM/DD/{streamId}/

Each stream folder contains: