convez.ai
MCP ServerIntegração & APIsIntermediário
MyMeetAI

MyMeet MCP Server

MyMeetAI/mymeet-mcp-server

Integre reuniões ao Claude com gravação, transcrição e análise automática

Ver no GitHub
InicialPluginScore · de 0 a 100

Servidor MCP que conecta seu assistente IA a reuniões em Google Meet, Zoom, Microsoft Teams e mais 6 plataformas. Oferece gravação, transcrição, busca, análise e exportação de reuniões diretamente do Claude, Cursor ou qualquer cliente compatível com MCP.

Instalação

$npx convez-ai MyMeetAI/mymeet-mcp-server

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
5
Forks
2
Acompanhando
0
Saúde
50%
Último commit
há 4 meses
Versão
—
Issues abertas
1
Contribuidores
4

Casos de uso

  • Transcrever e resumir reuniões automaticamente
  • Analisar conversas de vendas com templates de análise
  • Buscar informações em histórico de reuniões
  • Exportar transcrições e insights de reuniões

Veredito

MCP server robusto com suporte a múltiplas plataformas de conferência e dois modelos de deploy (local e remoto). Bem documentado e publicado no npm, atende bem equipes que precisam integrar reuniões ao fluxo do Claude — mantém-se focado e não promete milagres de IA.

README

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

MyMeet MCP Server

English | Русский

Connect your AI assistant to your meetings. Record, transcribe, search, analyze, and export meetings from Google Meet, Zoom, Microsoft Teams and 5 more platforms — directly from Claude, Cursor, Codex, or any MCP-compatible client.

  • 🧰 11 tools — list, search, check status, summarize, transcript, download, record, rename, re-analyze, edit, delete
  • 🗂️ 11 analysis templates — sales, HR, 1:1, research, protocol, medical, and more
  • 🔌 Two ways to run — locally over stdio (npx, zero config) or against a remote server over HTTP
  • 🔑 Simple auth — your MyMeet API key (an env var locally, Authorization: Bearer … over HTTP)
  • 📦 Published as @mymeet/mcp-server · Node ≥ 18 · built on the official MCP SDK

Need a key first? Get your API key at app.mymeet.ai/settings (or email hello@mymeet.ai for B2B access), then pick a setup below.


Which setup should I use?

Local — npm / stdioRemote — HTTP
Installnpx, runs on your machinenothing — just point at a URL
Best forClaude Desktop, Claude Code, Cursor, CodexTeam/hosted setups, browser clients
AuthMYMEET_API_KEY env varAuthorization: Bearer <key>
Hostingn/ause hosted mcp.mymeet.ai or self-host

Both expose the exact same tools — choose whatever your client supports. Using a client that isn't listed below? Jump to Universal setup.


Option 1 — Local (npm, stdio)

The server runs on your machine through npx; your API key lives in the client config and never leaves it. Requires Node 18+.

Claude Desktop

Easiest — the .mcpb bundle (one-click). Grab mymeet.mcpb from the releases page (or build it yourself with npm run pack:mcpb), then in Claude Desktop open Settings → Extensions → Advanced → Install Extension…, select the file, and paste your API key into the field — no JSON editing, and the key is stored in your OS keychain.

Or configure it manually via npx — edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "mymeet": {
      "command": "npx",
      "args": ["-y", "@mymeet/mcp-server"],
      "env": { "MYMEET_API_KEY": "your-api-key-here" }
    }
  }
}

Claude Code

claude mcp add mymeet --transport stdio -e MYMEET_API_KEY=your-key -- npx -y @mymeet/mcp-server

Cursor

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "mymeet": {
      "command": "npx",
      "args": ["-y", "@mymeet/mcp-server"],
      "env": { "MYMEET_API_KEY": "your-api-key-here" }
    }
  }
}

Codex (OpenAI)

codex mcp add mymeet --env MYMEET_API_KEY=your-key -- npx -y @mymeet/mcp-server

Or add the server to ~/.codex/config.toml by hand:

[mcp_servers.mymeet]
command = "npx"
args = ["-y", "@mymeet/mcp-server"]
env = { MYMEET_API_KEY = "your-api-key-here" }

Run /mcp inside Codex to confirm the server connected and see its tools.

Some other client? Every stdio MCP client reduces to the same three values — see Universal setup below.


Option 2 — Remote (HTTP)

No local install — connect your client to the server URL and authenticate per request. Great for browser-based clients and shared team deployments.

Hosted URL: https://mcp.mymeet.ai/mcp

Claude Desktop

Claude Desktop's JSON config works reliably through mcp-remote, which runs locally over stdio and forwards requests to the hosted MCP endpoint. Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "mymeet": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.mymeet.ai/mcp",
        "--header",
        "Authorization:${AUTH_HEADER}"
      ],
      "env": {
        "AUTH_HEADER": "Bearer your-api-key-here"
      }
    }
  }
}

Claude Code

claude mcp add mymeet --transport http https://mcp.mymeet.ai/mcp \
  --header "Authorization: Bearer YOUR_API_KEY"

Cursor

{
  "mcpServers": {
    "mymeet": {
      "url": "https://mcp.mymeet.ai/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}

Codex (OpenAI)

export MYMEET_API_KEY=YOUR_API_KEY
codex mcp add mymeet --url https://mcp.mymeet.ai/mcp --bearer-token-env-var MYMEET_API_KEY

Or add it to ~/.codex/config.toml directly:

[mcp_servers.mymeet]
url = "https://mcp.mymeet.ai/mcp"
bearer_token_env_var = "MYMEET_API_KEY"

Codex reads the token from the named env var and sends it as Authorization: Bearer … on every request, so the key stays out of the config file.


Universal setup (any MCP client)

Using a client that isn't listed above? Every MCP client reduces to the same handful of values — map them onto whatever shape it expects (a JSON mcpServers object, Codex's TOML [mcp_servers.*] table, or a … mcp add command). Both modes expose the identical tools.

SettingLocal — stdioRemote — HTTP
commandnpx—
args["-y", "@mymeet/mcp-server"]—
envMYMEET_API_KEY=<your key>—
url—https://mcp.mymeet.ai/mcp
auth—Authorization: Bearer <your key>

JSON clients (Claude Desktop, Cursor, VS Code, Cline, Windsurf, Zed, …) — local (stdio) mode:

{
  "mcpServers": {
    "mymeet": {
      "command": "npx",
      "args": ["-y", "@mymeet/mcp-server"],
      "env": { "MYMEET_API_KEY": "your-api-key-here" }
    }
  }
}

For remote (HTTP) mode, clients that support Streamable HTTP directly can swap command/args/env for a url and an auth header:

{
  "mcpServers": {
    "mymeet": {
      "url": "https://mcp.mymeet.ai/mcp",
      "headers": { "Authorization": "Bearer your-api-key-here" }
    }
  }
}

Claude Desktop can use the hosted endpoint through mcp-remote:

{
  "mcpServers": {
    "mymeet": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.mymeet.ai/mcp",
        "--header",
        "Authorization:${AUTH_HEADER}"
      ],
      "env": {
        "AUTH_HEADER": "Bearer your-api-key-here"
      }
    }
  }
}

TOML clients (Codex) — use the [mcp_servers.mymeet] blocks shown in Option 1 and Option 2.


Tools

Read-only tools default to the current user's meetings (scope: "mine"); pass scope: "workspace" to span the whole workspace (owners/admins). Full parameters, responses, and edge cases live in docs/TOOLS.md.

ToolTypeDescription
mymeet_list_meetingsreadList meetings (paginated). scope: "workspace" for all workspace meetings
mymeet_get_meeting_statusreadProcessing status: new → queued → processing → processed / failed
mymeet_get_meeting_reportreadAI summary: key points, action items, decisions (no transcript)
mymeet_get_transcriptreadFull transcript with speaker labels and timestamps
mymeet_search_meetingsreadSearch across all pages by title, people, date range, or status
mymeet_download_meetingreadExport report — md/json inline, pdf/docx as a download URL
mymeet_record_meetingwriteSchedule/start recording on 8 platforms, with optional cron
mymeet_rename_meetingwriteRename a meeting
mymeet_regenerate_templatewriteRe-analyze a meeting with a different template
mymeet_update_summarywriteEdit AI-generated summary sections
mymeet_delete_meetingwrite⚠️ Permanently delete a meeting

Resource: mymeet://templates returns the list of templates with descriptions, so the assistant can suggest the right one.


Templates

Used by mymeet_record_meeting and mymeet_regenerate_template:

TemplateUse case
default-meetingStandard summary with key points and action items
sales-meetingSales call: objections, next steps, deal signals
sales-coachingSales coaching feedback: technique, improvement areas
hr-interviewCandidate evaluation: strengths, concerns, key answers
research-interviewUser research: insights, patterns, methodology notes
team-syncPer-person updates, blockers, decisions
articleSEO article/blog post from meeting content
lecture-notesKey concepts, examples, study takeaways
one-to-oneManager 1:1: feedback, goals, action items
protocolFormal protocol: agenda, decisions, responsible parties
medicineMedical consultation: anamnesis, symptoms, recommendations

Supported platforms

Google Meet · Zoom · Microsoft Teams · Yandex Telemost · SberJazz · TrueConf · KonturTalk · Jitsi

The platform is auto-detected from the meeting URL, or set it explicitly with the source parameter.


Example prompts

"Show me my recent meetings"
"What was discussed in my last sales call?"
"Find my meetings with Vladimir from last week"
"Show all workspace meetings that failed processing"
"Record my Zoom meeting tomorrow at 2pm as a sales meeting"
"Re-analyze meeting X using the hr-interview template"
"Download the report for meeting Y as markdown"

Configuration

All configuration is via environment variables.

VariableModeRequiredDefaultDescription
MYMEET_API_KEYstdio✅ (stdio)—Your API key. Used in local/stdio mode.
MYMEET_ENABLE_SEARCH_TOOLboth—trueSet to false/0 to hide mymeet_search_meetings.
MYMEET_API_URLboth—https://backend.mymeet.aiOverride the backend base URL (dev/staging).
PORThttp—3000HTTP listen port (Railway sets this automatically).

Auth: In stdio mode the key comes from MYMEET_API_KEY. In HTTP mode each request carries its own API key in Authorization: Bearer … — there is no env-key fallback in HTTP mode, so every client presents its own key.


Self-hosting the remote server

The remote server is the same binary started with --http. It exposes:

PathPurpose
POST /mcpMCP Streamable HTTP endpoint
GET /healthHealth check (returns ok)

Docker

docker build -t mymeet-mcp .
docker run -p 3000:3000 mymeet-mcp
# Clients connect to http://localhost:3000/mcp with an Authorization header

The image runs node dist/index.js --http --port 3000 and ships a /health healthcheck.

Railway

railway.json is preconfigured (NIXPACKS build, npm start, /health healthcheck, restart-on-failure). Push the repo and set env vars in the Railway dashboard.

systemd + nginx (bare VM)

Reference units are in deploy/:

  • deploy/mymeet-mcp.service — runs node dist/index.js --http --port 3100 under systemd.
  • deploy/nginx-mcp.conf — reverse-proxies mcp.mymeet.ai → 127.0.0.1:3100. SSE-critical settings (proxy_buffering off, long read timeouts) are already in place for Streamable HTTP.
sudo cp deploy/mymeet-mcp.service /etc/systemd/system/
sudo systemctl enable --now mymeet-mcp
sudo cp deploy/nginx-mcp.conf /etc/nginx/sites-available/mcp && \
  sudo ln -s /etc/nginx/sites-available/mcp /etc/nginx/sites-enabled/ && \
  sudo nginx -s reload
# then issue TLS with certbot (see the commented lines in the conf)

How it works

MCP client (Claude / Cursor / Codex / Code / …)
        │  JSON-RPC over stdio  ── or ──  Streamable HTTP
        ▼
   index.ts      parse args, resolve credential, start transport
        ▼
   server.ts     register 11 tools + templates resource
        ▼
   client.ts     native fetch · 15s timeout · retry w/ backoff
        ▼
   https://backend.mymeet.ai/api/   (MyMeet REST API)

Design notes (logging that's safe for stdio, the report/transcript split, client-side search) are documented in docs/ARCHITECTURE.md.


Development

git clone https://github.com/MyMeetAI/mymeet-mcp-server.git
cd mymeet-mcp-server
npm install
cp .env.example .env   # add your MYMEET_API_KEY
CommandDescription
npm run devRun from source with tsx (hot reload, stdio)
npm startRun the built server in HTTP mode
npm run buildBundle to dist/index.js with tsup
npm testRun the test suite (vitest)
npm run test:watchTests in watch mode
npm run lintType-check with tsc --noEmit
npm run pack:mcpbBuild and package the .mcpb desktop bundle (→ mymeet.mcpb)

To run the remote server locally: npm run build && node dist/index.js --http --port 3000.

To build the desktop bundle: npm run pack:mcpb produces mymeet.mcpb (manifest + built server + production node_modules), installable via Claude Desktop's Install Extension….

License

MIT

Ver o original no GitHub

Bom para

Relacionados