Air-Gapped Local · Cloud-Ready · Privacy by Default

Your AI Co-Pilot
for Every Meeting.

Meem8 listens to your meetings and puts Gemma 4 on standby. Say the trigger word, get an instant answer — locally on your machine, or via a cloud bot in the call. You choose the mode based on what this meeting deserves.

Open DashboardCompare Modes

Privacy by Default

Starts in Air-Gapped Local Mode. Your microphone feeds directly into Gemma 4 running on your own hardware — no audio, transcript, or prompt ever touches a cloud server.

Always Listening, Never Intrusive

Transcripts run continuously in the background. Gemma 4 activates only when you say the trigger word — "hi gemma" — so the model stays silent until you actually need it.

Dual-Mode Architecture

Switch between Air-Gapped Local (browser mic, zero cloud) and Cloud Bot (Recall.ai joins the call for you) in a single toggle. Same Gemma 4 engine, different threat models.

Dual-Mode Architecture

Choose Your Mode

Not every meeting carries the same stakes. Meem8 adapts to the level of sensitivity the meeting demands.

Default

Air-Gapped Local

For Sensitive Meetings

Zero bytes leave your device. Your browser microphone feeds directly into a local Gemma 4 model via Ollama. No API calls, no cloud logs, no exposure — even to Meem8's own servers.

  • Legal & compliance reviews
  • HR, executive, and board discussions
  • M&A due diligence calls
  • NDA-protected client sessions
  • Anything you wouldn't want in a cloud log

Requires: Ollama + gemma4:latest

Browser: Chrome or Edge (Web Speech API)

Cloud Bot

For Routine Meetings

A Recall.ai bot joins your Zoom or Google Meet remotely — no microphone permissions, no screen sharing, no host requirements. Paste the link, invite the bot, and Gemma 4 handles the rest.

  • Daily standups and sprint reviews
  • Product demos and customer onboarding
  • Team all-hands and retrospectives
  • Multi-participant calls you're facilitating
  • Any meeting where convenience wins

Requires: Recall.ai API key

Requires: ngrok tunnel active (npm run tunnel)

Same Gemma 4 engine. Same trigger word. Same RAG pipeline.

Both modes feed into the same local Gemma 4 inference engine with full RAG context from your knowledge base. The only difference is how the audio gets in.

Try it now

Quick Start

Pick a mode and follow the steps.

01

Pull the model

ollama run gemma4:latest
02

Launch Meem8

npm run dev
03

Open the dashboard

http://localhost:3008/dashboard
04

Start secure mic

Click "Start Local Secure Microphone" — say "hi gemma" to query
Open Dashboard