meeting-recorder

Join Google Meet calls, transcribe audio in real-time, and participate via chat. Use when asked to join a meeting, transcribe a call, attend a video conference, or take meeting notes.

⭐2stars🔀0forks📁View on GitHub🕐Updated Dec 31, 2025⬇️3downloads

When & Why to Use This Skill

The Meeting Recorder skill empowers Claude to autonomously join Google Meet calls, perform real-time audio transcription, and participate in discussions via chat. By automating the documentation process and capturing timestamped transcripts, it transforms video conferences into searchable, actionable data, allowing users to focus on the conversation while the AI handles the note-taking.

Use Cases

  • Automated Meeting Minutes: Join Google Meet sessions to generate comprehensive, timestamped transcripts and summaries, eliminating the need for manual note-taking.
  • Real-time AI Interaction: Monitor live meeting audio for specific keywords or direct questions, enabling Claude to provide immediate information or responses via the meeting chat interface.
  • Knowledge Management: Create a local, searchable archive of past meetings including transcripts, metadata, and detected mentions for easy reference and team alignment.
  • Accessibility and Inclusion: Provide real-time text-based transcription for participants who may have hearing impairments or are joining from noisy environments.
namemeeting-recorder
descriptionJoin Google Meet calls, transcribe audio in real-time, and participate via chat. Use when asked to join a meeting, transcribe a call, attend a video conference, or take meeting notes.

Meeting Recorder

Join Google Meet calls as an active participant with real-time transcription.

Prerequisites

  1. Chrome automation skill must be installed and working
  2. Run setup once: ~/.claude/skills/meeting-recorder/scripts/meeting-recorder-setup/setup.sh

Quick Start

Join a Meeting

meeting-recorder join "https://meet.google.com/xxx-yyyy-zzz"

Check Status

meeting-recorder status

Read Live Transcript

# Current meeting
tail -f /tmp/meetings/current/transcript.txt

# Specific meeting
cat /tmp/meetings/abc-defg-hij/transcript.txt

Send Chat Message

meeting-recorder chat "Hello from Claude!"

Leave Meeting

meeting-recorder leave

List Past Meetings

ls /tmp/meetings/

Meeting Storage

Each meeting creates a directory at /tmp/meetings/<meeting-id>/:

  • transcript.txt - Full transcript with timestamps
  • metadata.json - Meeting info (URL, start time, status)
  • mentions.txt - Detected questions/mentions for Claude

The /tmp/meetings/current symlink always points to the active meeting.

Configuration

Edit ~/.meeting-recorder.json:

{
    "participant_name": "Claude Assistant",
    "meetings_dir": "/tmp/meetings",
    "mention_keywords": ["claude", "assistant", "ai"],
    "speaches_url": "ws://localhost:8000/v1/realtime",
    "transcription_model": "Systran/faster-distil-whisper-small.en"
}

How It Works

  1. Chrome navigates to meeting URL
  2. Enters participant name and joins (camera/mic off)
  3. PulseAudio captures Chrome's audio output
  4. Audio is batched (5-second chunks) and sent to Speaches HTTP API
  5. Transcription written to meeting directory
  6. Claude can read transcript and respond via chat

Active Participant Mode

Claude can monitor the transcript for mentions and questions:

  • Check /tmp/meetings/current/mentions.txt for detected questions
  • Respond via meeting-recorder chat "response"

For detailed documentation, see REFERENCE.md.