convez.ai
MCP ServerProdutividadeIntermediário
abhishekjoshi-eng

AbhiMeet - Gravador de Reuniões com IA Local

abhishekjoshi-eng/AbhiMeet

Grave reuniões localmente, transcreva com Whisper, relatorios com Claude MCP

Ver no GitHub
InicialPluginScore · de 0 a 100

Gravador de reuniões open-source que captura áudio e tela, transcreve localmente com Whisper AI em 90+ idiomas e gera relatórios via Claude MCP. Totalmente gratuito, sem nuvem, sem inscrição — alternativa local ao Fireflies.ai.

Instalação

$npx convez-ai abhishekjoshi-eng/AbhiMeet

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
1
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 7 meses
Versão
—
Issues abertas
1
Contribuidores
1

Casos de uso

  • Gravar e transcrever reuniões de trabalho sem depender de serviços cloud
  • Gerar automaticamente atas, action items e sumários via Claude
  • Analisar conteúdo visual de reuniões com extração de frames de vídeo
  • Documentar reuniões em idiomas não-ingleses (90+ suportados)

Veredito

Ferramenta sólida para equipes que valorizam privacidade e soberania de dados. A integração com Claude MCP é robusta (15 tools), mas requer Node.js + Python local; ideal para desenvolvedores e não-técnicos que aceitam setup inicial. Preenche lacuna real vs. Fireflies.ai.

README

A tradução para o português deste README ainda está na fila.


Why AbhiMeet?

FeatureFireflies.aiAbhiMeet
Monthly cost$19-39/userFree forever
Data storageTheir cloudYour local disk
PrivacyUploads to cloud100% offline
LanguagesLimited90+ via Whisper
AI ReportsBuilt-in (limited)Claude AI (unlimited)
Screen recordingNoYes
Custom reportsNoYes — MOM, action items, decisions, anything
Open sourceNoYes — MIT License

Demo

You:     "Transcribe my latest recording"
Claude:  Transcribing Budget-Review_160426_0230PM... Detected Hindi (92% confidence)
         Transcribed 47 segments in 12 seconds. Saved.

You:     "Generate meeting minutes with action items"
Claude:  ## Meeting Minutes — Budget Review (16 Apr 2026)
         ### Action Items:
         1. Abhishek to finalize vendor quotes by Friday
         2. Send revised budget to accounts team
         ...

Features

  • One-Click Recording — Single button to record mic + system audio + screen simultaneously
  • 3 Output Files Per Recording:
    • Subject_Audio_DDMMYY_HHMMam.mp3 — Audio only (for Claude transcription)
    • Subject_Video_DDMMYY_HHMMam.webm — Screen only (for visual analysis)
    • Subject_AV_DDMMYY_HHMMam.mp4 — Combined audio + video (for sharing)
  • Live Audio Waveform — Real-time visualization with frequency bars so you know recording is working
  • Recording Feedback — Animated indicators: "Audio capturing", "Screen capturing", "Saving to disk"
  • Built-in Media Player — Audio, Video, and Combined tabs with native playback
  • Whisper AI Transcription — Local speech-to-text (no API key, no internet required)
  • Video Frame Extraction — Claude can "see" what was on screen during the meeting
  • 15 MCP Tools — Full Claude integration for transcription, summaries, reports
  • Multi-Language — English, Hindi, Gujarati, Marathi, Bengali, Kutchi + 90 more
  • Configurable Storage — Browse and set your own recording folder
  • Dark Theme UI — Clean, professional, distraction-free interface
  • Zero Subscription — No cloud, no account, no recurring payments

Installation

Prerequisites

Quick Setup

# 1. Clone
git clone https://github.com/abhishekjoshi-eng/AbhiMeet.git
cd AbhiMeet

# 2. Install the Electron app
cd app
npm install
cd ..

# 3. Install the MCP server + Whisper
cd mcp-server
uv sync
cd ..

# 4. Download FFmpeg (Windows)
# Download from https://www.gyan.dev/ffmpeg/builds/
# Place ffmpeg.exe and ffprobe.exe in app/ffmpeg/

# 5. Run!
cd app
npx electron .

Connect to Claude Desktop (Cowork)

Add this to %APPDATA%/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "abhimeet": {
      "command": "uv",
      "args": ["--directory", "C:/path/to/AbhiMeet/mcp-server", "run", "main.py"]
    }
  }
}

Connect to Claude Code

Add this to .claude/.mcp.json:

{
  "mcpServers": {
    "abhimeet": {
      "command": "uv",
      "args": ["--directory", "/path/to/AbhiMeet/mcp-server", "run", "main.py"]
    }
  }
}

Restart Claude. You'll see 15 AbhiMeet tools available.

MCP Tools (15)

ToolWhat It Does
list_recordingsList all meeting recordings with metadata
get_recording_infoGet full details about a specific recording
get_audio_file_pathGet path to the audio MP3 for analysis
get_video_file_pathGet path to the screen recording WebM
get_combined_file_pathGet path to the combined A+V MP4
transcribe_recordingTranscribe audio using Whisper AI (runs locally, no API key)
extract_video_framesExtract screen frames so Claude can see what was shown
save_transcriptionSave transcription text to the recording folder
read_transcriptionRead back a saved transcription
save_summarySave a meeting summary
read_summaryRead a saved summary
save_reportSave reports — MOM, action items, decisions, follow-ups
search_recordingsFull-text search across all recordings
get_storage_statsStorage usage and disk space stats
delete_recordingDelete a recording and all its files

How It Works

Record Meeting          Whisper Transcribes        Claude Analyzes
    |                        |                         |
    v                        v                         v
[AbhiMeet App]  --->  [MCP Server]  --->  [Claude Desktop/Code]
    |                        |                         |
    v                        v                         v
3 files saved         Speech -> Text            Reports, MOM,
(MP3+WebM+MP4)     (90+ languages)          Action Items, Summary

Usage Examples

After connecting to Claude, just talk naturally:

What You SayWhat Happens
"List my recordings"Shows all recordings with dates, durations, file types
"Transcribe my latest recording"Whisper runs locally, saves transcription
"Generate meeting minutes"Claude reads transcription, creates structured MOM
"What action items came up?"Extracts and lists all commitments and deadlines
"Summarize in Hindi"Generates summary in Hindi from English recording
"What was shown on screen?"Extracts video frames, Claude describes visual content
"Search recordings for 'budget'"Finds all recordings mentioning budget

File Naming Convention

Every recording creates a folder with 3 properly named files:

Budget-Review_160426_0230PM/
    Budget-Review_Audio_160426_0230PM.mp3      (Audio only)
    Budget-Review_Video_160426_0230PM.webm     (Screen only)
    Budget-Review_AV_160426_0230PM.mp4         (Combined)
    metadata.json                              (Recording metadata)

Format: Subject_Type_DDMMYY_HHMMam.ext

Supported Languages

Powered by Whisper AI — supports 90+ languages including:

LanguageCodeLanguageCode
EnglishenHindihi
GujaratiguMarathimr
BengalibnTamilta
TeluguteKannadakn
UrduurPunjabipa
ArabicarChinesezh
JapanesejaKoreanko
FrenchfrGermande
SpanishesPortuguesept

Project Structure

AbhiMeet/
  app/                          # Electron desktop app
    src/
      main/                     # Main process
        index.js                # App window, permissions, tray
        ipc-handlers.js         # Recording logic, 3-file output
        file-manager.js         # FFmpeg conversion, file naming
        settings.js             # Settings + MCP config sync
        tray.js                 # System tray
      renderer/                 # Frontend UI
        index.html              # Dark theme, waveform, player
        styles.css              # Professional dark UI
        app.js                  # Recording, playback, waveform
      preload.js                # Secure IPC bridge
    assets/icon.svg             # App icon
    ffmpeg/                     # FFmpeg binaries (not in repo)
    package.json
  mcp-server/                   # Python MCP server
    main.py                     # 15 tools + Whisper transcription
    config.json                 # Storage path config
    pyproject.toml              # Dependencies
  recordings/                   # Auto-created recording storage
  LICENSE                       # MIT License
  README.md

Tech Stack

ComponentTechnologyPurpose
Desktop AppElectron 33Cross-platform recording UI
Audio/VideoFFmpegMP3 conversion, A+V merging
Transcriptionfaster-whisperLocal Whisper AI (no cloud)
AI IntegrationMCP (Model Context Protocol)Claude tools
MCP ServerFastMCP (Python)15-tool server
UIHTML/CSS/JSDark theme with live waveform

Contributing

Contributions are welcome! Feel free to:

  • Open issues for bugs or feature requests
  • Submit pull requests
  • Add support for more platforms (macOS, Linux)
  • Improve Whisper model selection
  • Add speaker diarization

Roadmap

  • macOS and Linux support
  • Speaker diarization (who said what)
  • Auto-record when Zoom/Meet/Teams call starts
  • Calendar integration for auto-naming
  • Export reports to PDF/Word
  • Electron auto-updater

License

MIT License - see LICENSE for details.

Author

Abhishek Tansukh Joshi
GitHub: @abhishekjoshi-eng


Ver o original no GitHub

Bom para

Relacionados