MeetStream MCP Server

Give Claude - or any MCP-compatible client - direct access to the MeetStream meeting-bot API. The server authenticates to MeetStream with your API key and exposes the full bot lifecycle (create, record, transcribe, summarize, interact live, manage calendar auto-join) as 19 callable tools.

Two ways to run it:

• Local (stdio) - the default, npx @meetstream/mcp. Your MCP client launches it as a subprocess; nothing to host.
• Remote (Streamable HTTP) - https://mcp.meetstream.ai/mcp. A hosted, multi-tenant endpoint you can add by URL, no local install. Each request authenticates with its own API key via header (see Remote server below).

Local:   Claude/MCP client ──stdio (subprocess)──► @meetstream/mcp ──HTTPS + API key──► api.meetstream.ai
Remote:  Claude/MCP client ──HTTPS + your key───► mcp.meetstream.ai ──HTTPS + your key──► api.meetstream.ai

1. Get an API key

Create one at app.meetstream.ai/api-keys. This is the credential the server uses to authenticate every request - internally it sends Authorization: Token <your-key> on every call to https://api.meetstream.ai/api/v1. There's no OAuth flow, no login step, no separate MCP account: the API key is the authentication.

2. Install & configure

The server reads the key from the MEETSTREAM_API_KEY environment variable - you set it once in your MCP client's config and every tool call uses it automatically.

Claude Code (CLI)

claude mcp add meetstream --env MEETSTREAM_API_KEY=ms_XXXX -- npx -y @meetstream/mcp

Claude Desktop

Edit claude_desktop_config.json (Settings → Developer → Edit Config):

{
  "mcpServers": {
    "meetstream": {
      "command": "npx",
      "args": ["-y", "@meetstream/mcp"],
      "env": { "MEETSTREAM_API_KEY": "ms_XXXX" }
    }
  }
}

Restart Claude Desktop after saving.

Cursor

Settings → Cursor Settings → MCP → Add new global MCP server, or edit ~/.cursor/mcp.json:

{
  "mcpServers": {
    "meetstream": {
      "command": "npx",
      "args": ["-y", "@meetstream/mcp"],
      "env": { "MEETSTREAM_API_KEY": "ms_XXXX" }
    }
  }
}

Windsurf

Settings (Cmd+,) → Cascade → MCP Servers → Add Server - same command/args/env as above, written to ~/.codeium/windsurf/mcp_config.json.

Any other MCP client

Point it at the stdio command npx -y @meetstream/mcp with MEETSTREAM_API_KEY set in the process environment. That's the entire integration surface - the server needs no other config.

Verify it's connected

Ask your client something like "list my meetstream bots" - if it returns data (or a clean empty list) instead of an error, auth is working. A misconfigured key surfaces as a tool error: "MEETSTREAM_API_KEY is not set..." or a 401 from the API.

Then just ask

"Join my standup at https://meet.google.com/abc-defg-hij, record it with Deepgram transcription, and give me the transcript when it's done."

"List my bots from today and give me the AI summary of the 3pm customer call."

"Send 'we'll follow up by email' into the meeting the bot is in, then show them our logo."

"Schedule a bot for tomorrow's board meeting on my calendar."

Full capability list - 19 tools

Bot lifecycle

Transcription

Media & meeting data

Live meeting interaction

Calendar

Reference

What makes this different from just reading the docs

Every tool description and the webhook_events_guide bake in live-verified ground truth, confirmed against real production bot runs, that the public API docs currently get wrong:

• Webhook envelope key is event, not bot_event as the docs claim.
• bot.stopped is two-layer: it fires exactly once, and bot_status (Stopped/NotAllowed/Denied/Error) tells you why - there's no separate bot.kicked/bot.denied event.
• Streaming-only transcription providers (deepgram_streaming, assemblyai_streaming, meeting_captions) never fire transcription.processed or bot.done - their terminal event is audio.processed. A handler waiting on bot.done for a streaming bot will hang forever.
• transcript_id is never in a webhook payload - get_transcript resolves it for you automatically instead of making the model guess.
• Safe defaults everywhere: automatic_leave timeouts on every create_bot call, and recording_permission_denied_timeout floored at 60 (the API rejects lower values with a 400).

Configuration

Remote server (Streamable HTTP)

https://mcp.meetstream.ai/mcp - a hosted, multi-tenant Streamable HTTP endpoint. No npx, no local Node, no per-machine install. Add it by URL:

{
  "mcpServers": {
    "meetstream": {
      "url": "https://mcp.meetstream.ai/mcp",
      "headers": { "Authorization": "Bearer ms_YOUR_API_KEY" }
    }
  }
}

How auth works here is different from stdio mode: this endpoint serves many different MeetStream accounts at once, so it holds no API key of its own. Every request must carry your key, either as Authorization: Bearer <key> or X-MeetStream-Api-Key: <key>. A request with no key gets a 401 with setup instructions instead of silently failing.

The server is stateless - every request is independent, there's no session to keep alive, and it scales horizontally with zero shared state between requests.

Self-hosting it yourself? The same code ships as a Docker image - see deploy/ for the full runbook (fresh isolated VM, nginx, Let's Encrypt, systemd) or just:

docker build -t meetstream-mcp .
docker run -p 8080:8080 meetstream-mcp   # POST http://localhost:8080/mcp

Troubleshooting

Development

npm install
npm test        # spawns the real stdio server and speaks JSON-RPC to it end-to-end

Prefer a terminal? See the MeetStream CLI. Docs: docs.meetstream.ai · API spec: openapi.json · Migrating from Recall.ai: @meetstream/migrate

MIT © MeetStream.ai

Telemetry

This server sends anonymous usage events (which tools are called, remote vs local) to help us improve MeetStream. It never sends your API key, meeting URLs, transcripts, or any content. Disable it any time:

export MEETSTREAM_TELEMETRY=0   # or the standard DO_NOT_TRACK=1