convez.ai
MCP ServerIntegração & APIsIntermediário
Percona-Lab

ECHO — Buscador de Transcrições de Reuniões Zoom

Percona-Lab/ECHO

Acesse suas transcrições Zoom via AI com segurança total

Ver no GitHub
InicialPluginScore · de 0 a 100

MCP server que integra transcrições de reuniões Zoom ao Claude e outras ferramentas MCP. Permite buscar, recuperar e resumir suas gravações de encontros diretamente via comandos de slash, usando OAuth 2.0 seguro sem armazenar credenciais localmente.

Instalação

$npx convez-ai Percona-Lab/ECHO

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
1
Acompanhando
0
Saúde
25%
Último commit
há 4 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Buscar informações específicas em transcrições de reuniões passadas
  • Gerar resumos automáticos de calls e hearings
  • Integrar histórico de reuniões ao fluxo de trabalho com IA

Veredito

Solução prática para quem trabalha com Zoom e quer acessar transcrições via Claude. Implementação segura com OAuth/PKCE, mas limitada a dados próprios (host-only) e dependente de permissões administrativas no Zoom. Útil em contexto corporativo com Zoom estabelecido.

README

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

ECHO - Explore Calls, Hearings & Observations

✅ Development restarted (June 2026)

ECHO now supports Zoom AI Companion Call Notes. Zoom exposes a user-level OAuth scope for AI Companion meeting summaries (meeting:read:summary), so ECHO can retrieve meeting summaries even when cloud recording was not active. One Zoom limitation remains: summaries are host-only at user scope, so you can retrieve notes from meetings you hosted, not ones you merely attended.

Rollout is pending two user-managed scopes (meeting:read:summary, meeting:read:list_meetings) being added to your org's Zoom OAuth app, after which each user re-runs echo-login once.

MCP server for searching your Zoom meeting transcripts from any MCP-compatible AI tool. Part of the Alpine Toolkit.

Uses OAuth 2.0 + PKCE so no secrets ever touch your machine. You log in with your own Zoom account and ECHO can only see your recordings.

Install

curl -fsSL https://raw.githubusercontent.com/Percona-Lab/ECHO/main/install.sh | bash

The installer walks you through:

  1. Cloning ECHO and installing its Python deps via uv
  2. Looking up your org in the public registry (for Percona, no further questions — just enter percona)
  3. Registering ECHO with Claude Desktop, Claude Code, or whatever MCP client you have
  4. One-time Zoom OAuth login

Once installed, your AI tool gets these slash commands automatically: /echo_status, /echo_recent, /echo_search <query>, /echo_summary <id>, /echo_transcript <id>, /echo_notes <meeting>.

The installer will guide you through:

  1. Cloning the repo and installing dependencies
  2. Configuring your Zoom OAuth Client ID (or using the registry)
  3. Authenticating with your Zoom account
  4. Auto-detecting and configuring your AI client

Prerequisites

You need a Zoom OAuth Client ID from a General App (OAuth 2.0). If you have admin or owner access to your Zoom account, you can create this yourself. If not, ask your Zoom admin.

Creating a Zoom OAuth App

  1. Go to marketplace.zoom.us > Develop > Build App
  2. Select General App (OAuth 2.0)
  3. Set redirect URL to http://localhost:8090/callback
  4. Under Scopes, choose User-managed and add:
    • cloud_recording:read:content
    • cloud_recording:read:list_user_recordings
    • meeting:read:summary
    • meeting:read:list_meetings
    • user:read:user
  5. Activate the app
  6. Copy the Client ID (you do not need the Client Secret)

The PKCE flow means the Client Secret stays in the Zoom admin console and never needs to be shared.

Organization registry

If your org has already set up the Zoom OAuth app, the installer will find the Client ID automatically when you enter your Zoom subdomain (e.g. acme from acme.zoom.us).

To register your org's Client ID so others can skip this step, submit a PR adding your org to client_registry.json:

{
  "orgs": {
    "yourcompany": {
      "name": "Your Company",
      "client_id": "YOUR_CLIENT_ID_HERE"
    }
  }
}

Client IDs are public identifiers, safe to commit.

Tools

ToolDescription
auth_statusCheck if ECHO is connected to your Zoom account
list_meetingsList your recent meetings with cloud recordings
get_transcriptGet the full transcript for a specific meeting
search_transcriptsSearch across your meeting transcripts by keyword or phrase
meeting_summaryGet participants and a condensed conversation flow
list_past_meetingsList recent meetings you hosted, including ones without recordings
ai_meeting_notesGet the AI Companion (note-taking) summary for a meeting you hosted

CLI Commands

CommandDescription
uv run echo-loginAuthorize ECHO with your Zoom account
uv run echo-logoutRemove stored tokens
uv run echo-mcpRun the MCP server

Security Model

ECHO is designed so that no org-level secrets ever touch your machine.

WhatWhereWho controls it
Client SecretZoom admin consoleZoom admin (never shared)
Client IDYour .env fileYou (public identifier, safe to share)
OAuth tokens~/.echo/tokens.json (mode 600)You (scoped to your account only)

How it works:

  • You authenticate by signing into Zoom in your browser (OAuth + PKCE)
  • ECHO uses /users/me/ endpoints, so it can only see your recordings
  • Tokens auto-refresh so you stay logged in without re-authenticating

Manual Setup

If you prefer not to use the installer:

git clone https://github.com/Percona-Lab/ECHO.git
cd ECHO
uv sync
cp .env.example .env        # Add your Zoom Client ID
uv run echo-login            # Authenticate with Zoom

Then add to your MCP client config:

{
  "mcpServers": {
    "echo": {
      "type": "stdio",
      "command": "uv",
      "args": ["run", "--directory", "/path/to/ECHO", "echo-mcp"]
    }
  }
}

Development

uv sync
uv run echo-mcp                                          # Run server
npx @modelcontextprotocol/inspector uv run echo-mcp      # MCP Inspector

Built with

  • FastMCP - Python MCP framework
  • CAIRN - Alpine Toolkit scaffolder
Ver o original no GitHub

Bom para

Relacionados