TrainAR Docs
Try Free
AR Glasses (Core App) · Last updated 18 May 2026 · 4 min read

Core App overview

The Core App is the TrainAR software that runs on the AR glasses themselves — what your engineer wears on a job.

The Core App is the TrainAR software that runs on the AR glasses themselves — what your engineer wears on a job. It's a separate piece of software from the Dashboard (which is the office-side web app for managing your team and content).

Two-line summary:

  • Dashboard = web app for admins, training managers, content authors.
  • Core App = on-glasses software for the engineer doing the work.

If you've heard us talk about "the glasses app" or "the on-glasses experience," that's the Core App.

What it does

When an engineer puts on a paired pair of glasses and opens TrainAR, the Core App:

  1. Authenticates the device against your tenant.
  2. Loads the right user's seat, role (trainer vs trainee), and entitlements (which skills + knowledge they have access to via your bundle subscriptions).
  3. Connects a real-time voice channel to a TrainAR AI assistant (built on OpenAI's Realtime API).
  4. Starts streaming the engineer's point-of-view video for the session.
  5. Lets the engineer get on with the job — voice-driven, hands-free, eyes on the work.

When the session ends, the recorded video is processed and structured procedures are extracted, all back-on-the-platform side. The engineer doesn't have to do anything more.

What it isn't

  • Not a mobile app. There is no separate phone app to install. The Core App runs natively on the glasses.
  • Not the Dashboard. Admins manage seats, tasks, integrations, billing, skills, knowledge, and reports in the Dashboard — not on the glasses.
  • Not a desktop app. The Core App is on-glasses only.

The glasses + the Core App relationship

The Core App is built to work specifically with two AR glasses platforms:

  • INMO Air 3 (primary, recommended)
  • Vuzix Blade 2

Each is a full Android device in a glasses frame — the Core App runs natively on the device. See Supported AR glasses for the full spec comparison.

Two roles: trainer and trainee

Every user (every seat) is either a trainer or a trainee.

  • Trainers wear the glasses to record a procedure — capturing how an expert actually does the job, voice-narrated and POV-recorded. Their sessions feed your procedure library.
  • Trainees wear the glasses to follow procedures — guided through a job by the AI assistant, with the senior's procedure as the source of truth, plus the rest of your knowledge base on tap.

The role is set on the seat in the Dashboard (Users & Seats). Trainer and trainee modes use different AI configurations under the hood — trainer is a low-cost listening + transcription configuration; trainee is the full conversational AI with skill execution.

See Trainer mode and Trainee mode for what each role experiences in a session.

Voice triggers

There's no wake-word — the AI doesn't sit there always listening for "Hey TrainAR." Sessions are explicitly started and stopped with voice:

  • "Start training session" — opens the voice channel and begins the session.
  • "Stop training session" — closes it gracefully.

Inside the session, the AI is always listening — speak naturally and it responds with no command word needed.

See The AI assistant on the glasses for the full conversational model.

What runs in the cloud vs on the glasses

The Core App is a thin client. The heavy lifting happens in TrainAR's cloud:

  • The AI assistant runs in the cloud (OpenAI Realtime API), not on-device.
  • Procedure extraction from session video happens after the session ends, in the cloud (Gemini 2.5 Pro).
  • Knowledge base search runs against TrainAR's pgvector store, not on-device.
  • Skill execution (custom webhooks, Parts Arena lookups, etc.) runs on TrainAR edge functions.

The glasses themselves handle: video capture, voice in/out, the small on-glass display, and the WebRTC connection to the AI room. They need a connection (Wi-Fi or LTE-via-tether) to be useful.

Where to next