MeetStream Documentation

Every template lives in MeetStream Labs and runs on its own. No shared build, no framework to learn first: Node 18+, a .env file and one command.

Building with a coding agent? Point it at llms.txt, connect the MeetStream MCP server, and paste the guide from Agent Quickstarts. Or tell it to start from one of the templates below.

Run any template

Clone Labs

git clone https://github.com/meetstream-ai/labs.git
cd labs/quickstart-first-bot

Add your API key

cp .env.example .env   # set MEETSTREAM_API_KEY

Create a key at app.meetstream.ai. Each template’s .env.example lists every variable it reads, with comments.

Install and run

npm install
node index.js

Templates that receive webhooks need a public HTTPS URL. webhook-local-tunnel shows how to get one for your laptop.

Popular starting points

Your first bot

Send a bot, poll its status, print the result. The hello world.

AI notetaker that emails notes

Record, summarize and email the notes to every attendee.

Real-time transcription

Stream the transcript while the meeting is still running.

Voice agent in a meeting

Deploy a MIA agent that listens and talks back.

Bots from the calendar

A bot joins every meeting on a connected calendar.

Production webhook handler

Every event, deduped, with correct terminal handling.

Start here

Template	What it shows
`quickstart-first-bot`	The hello world: send a bot into a meeting, poll status, print the result
`post-call-transcription`	Bot joins, records, and saves the full transcript when the call ends
`transcript-fetcher`	The canonical transcript flow: resolve `transcript_id`, poll through HTTP 202, parse segments
`webhook-handler-complete`	The reference webhook receiver: every event, deduped redeliveries, correct `status_code` reading

Related guides: Create your first bot, Webhooks and events

Bots and lifecycle

Template	What it shows
`quickstart-first-bot`	Create a bot, poll `GET /bots/{id}/status` until the session ends
`bot-status-monitor`	Render the bot lifecycle as a live timeline in your terminal
`list-and-manage-bots`	List every bot, filter and sort, make an active bot leave
`bot-lifecycle-state-machine`	Webhook-driven state machine, persisted to disk, detects stuck and abandoned bots
`scheduled-bot-join-at`	Schedule with `join_at`, then reschedule and cancel
`manage-scheduled-bots`	Admin CLI for bots that have not joined yet
`bot-retention-config`	Control how long recordings live with `recording_config.retention`
`delete-bot-data`	Permanently erase a bot’s data, behind a typed confirmation
`idempotency-and-dedup`	Both anti-duplicate mechanisms: `Idempotency-Key` (507 replay) and `deduplication_key` (409)
`bulk-bot-operations`	Run many bots concurrently with a bounded queue and per-job idempotency

Recording and media

Template	What it shows
`audio-recording-downloader`	Wait for `audio.processed`, then download the audio
`video-recording-downloader`	Same for video, via `video.processed`
`per-participant-audio-recorder`	One audio file per participant, with correct HTTP 202 handling
`per-participant-video-recorder`	One video and audio stream per participant
`screenshot-capture`	Download meeting screenshots and place them on the timeline
`pause-resume-recording`	Open and close recording privacy windows mid-meeting
`custom-s3-storage`	Bring your own bucket: point MeetStream at your own S3 via `PUT /admin/configs`

Related guides: Per-participant video, Custom storage (S3)

Transcription

Template	What it shows
`post-call-transcription`	Full transcript saved after the call
`realtime-transcription`	Streaming transcription over webhook or WebSocket while the meeting runs
`transcript-fetcher`	Resolve `transcript_id` and fetch, handling 202 correctly
`re-transcribe-audio`	Re-run transcription with a different provider, model or language
`multi-provider-transcription`	All five post-call providers, switched by one env var
`multilingual-transcription`	Non-English meetings with the right provider and language code
`speaker-diarization`	Deepgram diarization into clean per-speaker turns
`live-captions-overlay`	Rolling live captions in your terminal

Meeting data and analytics

Template	What it shows
`ai-meeting-summary`	MeetStream’s native AI summary (`GET /bots/{id}/summary`)
`speaker-timeline-analytics`	Talk-time, turns, longest monologue, ASCII bar chart
`participant-tracker`	Who joined and left, live over webhooks
`meeting-chat-logger`	Export in-meeting chat to JSON and Markdown
`meeting-analytics-dashboard`	Everything about one meeting in a single consolidated report

Related guides: API reference

Live interaction

Template	What it shows
`realtime-audio-streaming`	Live PCM audio over a WebSocket you host
`realtime-video-streaming`	Live fMP4 video over WebSocket, with the ping/pong keepalive
`send-chat-message`	Post chat messages into a live meeting, on demand or scheduled
`send-image-bot`	Put an image or GIF into the chat, or make it the bot’s camera feed
`websocket-bot-control`	The bring-your-own bridge: `sendaudio`, `sendmsg`, `sendchat`, `interrupt`, `sendimg`
`interactive-meeting-agent`	Two-way loop: hear the room, decide, respond

Related guides: Meeting control patterns

MIA voice agents

Template	What it shows
`MIA-chat-agent`	Deploy a MIA into a meeting and watch it respond
`mia-voice-agent-pipeline`	Pipeline mode: swap STT, LLM and TTS layers independently
`mia-realtime-agent`	Realtime mode: a single speech-to-speech model, lower latency
`mia-agent-crud`	Admin CLI for saved agent configs
`mia-wake-word-assistant`	Only answers when addressed: “hey acme, …”

Related guides: What is MIA, Create an agent

Calendar

Template	What it shows
`google-calendar-integration`	Connect Google Calendar, including a local OAuth helper
`outlook-calendar-integration`	Connect Outlook / Microsoft 365, including the Azure setup
`calendar-event-sync`	Sync upcoming events and find the ones with a join link
`calendar-schedule-bot`	Schedule a bot for one event, handling the 409 duplicate case
`calendar-auto-schedule`	A bot joins every meeting automatically
`calendar-recurring-events`	Whole series or single occurrence, plus auto-rescheduling
`calendar-disconnect`	Tear down an integration, with a preview of what is lost

Related guides: Google Calendar setup, Outlook Calendar setup

Platform specifics

Template	What it shows
`zoom-meeting-bot`	Zoom’s recording-permission flow and its timeout ranges
`zoom-authenticated-joins`	Host the ZAK / OBF token URL so bots join Zoom as a signed-in user, not a guest
`teams-meeting-bot`	Microsoft Teams lifecycle and admission behaviour
`gmeet-lobby-handling`	Google Meet waiting room: detect `NotAllowed` and `Denied`, then react
`google-signed-in-bots-setup`	SAML SSO, certificates and domain registration, end to end
`google-login-management`	Admin CLI for signed-in bot domains and logins

Related guides: Zoom, Zoom authenticated bots, Google signed-in bots

Webhooks and reliability

Template	What it shows
`webhook-handler-complete`	Every documented event, deduped, with correct status handling
`webhook-local-tunnel`	Get webhooks delivering to your laptop over a public HTTPS tunnel
`error-handling-and-retries`	The whole error surface: 400, 401 vs 403, 409, 429 with `Retry-After`, 507, 202

Related guides: Webhooks and events, Local webhook server, Errors

Integrations and AI workflows

Template	What it shows
`ai-meeting-notetaker-email`	Record, summarize, and email the notes to attendees
`slack-meeting-summary`	Post the summary and action items into Slack
`notion-meeting-notes`	One Notion page per meeting
`crm-hubspot-sync`	Attach the summary and transcript to the HubSpot contact and deal
`action-item-extractor`	LLM-extracted action items with owners and due dates
`sentiment-and-insights`	Sentiment, topics and talk-time balance as a coaching report

Related guides: Agent Quickstarts

Agent and assistant integrations

Larger reference integrations that connect MeetStream MIA voice agents to other agent frameworks.

Project	What it shows
`mia-hermes-bridge`	Bridge a self-hosted Hermes Agent into a live meeting through MIA, with ngrok tunnelling and a one-command launcher
`MIA-poster-design`	A wake-word MIA assistant that turns a spoken event brief into a poster through the Canva bridge
`openclaw-x-meetstream-skill`	An OpenClaw skill for operating MeetStream bots from the OpenClaw agent

What every template gets right

The templates are written against behaviour verified on the live API, so they double as a reference for the details that most often go wrong:

REST auth is Authorization: Token <key>. The MCP server uses Bearer.
Terminals are two-layer. Every ending arrives once as event: "bot.stopped", and bot_event gives the reason: bot.stopped, bot.kicked, bot.notallowed, bot.denied or bot.failed. bot.done is the final event on every path.
Transcripts are fetched by transcript_id, and segments carry transcript, not text.
202 means not ready, 507 means an idempotent replay. Every poll loop has a cap, and a replay counts as success.
Streaming-only providers produce no post-call transcript. The live stream is the record, or re-transcribe the stored audio afterwards.
MIA bots take only agent_config_id.

Contribute a template

Have a use case the list is missing? Templates follow one layout (README.md, .env.example, package.json, .gitignore, index.js, src/*.js) and every endpoint they call is checked against the API before merge. Open a pull request on MeetStream Labs.