Key Concepts

Assistant

An Assistant is the template each live instance is cloned from the moment it joins a meeting. Think of it as the recipe that decides flavor, language, and personality.

🔬 Anatomy of an Assistant

Component	Why It Matters	Example
Prompt Template	Gives the LLM its mission.	“You are a cheerful scrum master…”
Transcription Engine	Decides how we hear people. Supports all Deepgram and OpenAI models	Deepgram Nova‑3 for lightning‑fast multilingual transcription.
Diarization	Who said what? Speaker separation for crystal-clear minutes.	true / false
Key Terms	Teaches the transcription engine brand‑new jargon - perfect for niche acronyms and uncommon proper nouns.	"Exlabs", "AcmeX", "Kubernetes"

Use‑case sneak peek

Launch an HR Interview Assistant that evaluates candidate answers in real‑time, flags missing competencies, and suggests laser‑focused follow‑up questions to the interviewer.

Meeting

A Meeting is a single-use instance of an Assistant. It represents self-contained execution of your meeting assistant logic within the context of a specific live call - from join to leave.

🎬 MeetingLifecycle at a Glance

🛠️ Key Settings

• Meeting URL – Google Meet, Teams, or any provider MCP surfaces.
• Prompt Parameters – e.g., { sprint = 42, team = "Alpha" } to customize Assistant per each meeting.
• Recording & Transcription Flags – toggle according to your needs.

Real‑world hook

A PM schedules a Meeting for every daily stand‑up; the Assistant records meetings detects blockers and lists action points from previous meetings.

Scheduler – Calendar Whisperer

Tired of cron jobs? Scheduler plugs directly into your calendars (Google, Outlook, or any iCal feed) and auto‑spawns Meetings right on schedule.

🔧 Key Parameters

Parameter	Purpose	Example
Enabled Calendars	Calendar IDs the Scheduler listens to.	"[email protected]", "[email protected]"
Event Name Pattern	Glob‑style filter that decides which events the Assistant joins. Use * to match all meetings.	"Weekly-Design-Review*"

Thread

Threads aren’t just for retrospectives-they stream real‑time context from the active Meeting. That means you can whisper questions to the Assistant and receive answers before the speaker finishes the next sentence.

Things you can do on the fly:

• Ask for a lightning summary of the last minute of dialogue.
• Check whether blockers or deadlines have already been mentioned.
• Instantly fire a Tool, e.g., “slack remind John to send the deck.”

After the meeting, the same Thread lives on for asynchronous follow‑ups-perfect for generating polished summaries, drafting emails, or nudging teammates.

Tools – Turning Words into Actions

Tools are the hands and legs of the platform.

🛠️ Built‑in Tools

Tool	What It Does	Splashy Example
Slack	Send or schedule channel messages, upload files, read messages.	Post AI minutes to #general 5 s after the call.
Jira	Create, read, assign, or transition issues.	Open a “Bug: onboarding flow crash” and assign to @alice.

🧩 Custom Tools (Zapier, N8N, your API)

Define a JSON schema, drop a webhook URL, decide on parameters, and voilà, any workflow is just a tool invocation away.

{
  "name": "triggerN8N",
  "description": "Start the build‑deploy pipeline",
  "parameters": {
    "type": "object",
    "properties": {
      "branch": { "type": "string" }
    },
    "required": ["branch"]
  },
  "server_url": "https://hooks.n8n.cloud/trigger"
}

Hook in action: After a QA call, the Assistant pipes “release candidate ready” to an n8n flow that spins up staging and pings QA on Slack - no human clicks required.

🤝 Model Context Protocol Integration

MCP (Model Context Protocol) is an open protocol that standardizes how applications provide context to LLMs. AI Meeting Assistant connects as an MCP Client, which means it can latch onto any MCP - instantly unlocking thousands of ready‑made integrations.

Need to surface product notes from Confluence mid‑discussion? Want to fire off a personalized follow‑up email the moment the call ends? If the MCP Server exposes it, the Meeting Assistant can reach it with a single tool invocation.

• Huge Integration Catalog – Tap into every connector the server offers - Salesforce, Confluence, ServiceNow, email gateways, and more.
• Fetch or Act – Pull knowledge bases into the meeting context or push outcomes out to downstream systems.
• Centralized Policy – The Voilo MCP broker still enforces access rules, rate limits, and audit trails.

The Big Picture

Diagram below illustrates how data and actions flow through the system: an Assistant blueprint becomes a live Meeting instance; the Thread enables real-time chat; the Scheduler handles auto-joining; and Tools allows to enable actions.

🚀 Why Stakeholders Love This

• Real‑time Answers – Ask “What did Alice just promise?” and get clarity instantly.
• Minutes in Minutes – Auto‑summaries land before attendees leave the room.
• Zero‑Touch Action Items – Jira issues materialize without toggling tabs.
• Insight at Scale – Harvest decisions, blockers, and sentiment across every meeting.
• Safe & Compliant – Deepgram/OpenAI models run in a secured, auditable pipeline.