FAQ · straight answers
The questions we actually get.
Answered straight.
No corporate fog: here’s what people ask us in demos, emails and at booths — with the answers we actually give. Every answer stands on its own and points to the page that digs deeper. Money questions? They have their own FAQ, on the pricing page.
The real questions we get about the product and about trust, with answers that stand on their own: what an AI agent is compared to a chatbot; where your data goes depending on the brain you pick; how a meeting gets captured with no clandestine recording; which model runs underneath and what Luge learns — or doesn't — from your data. Money questions have their own FAQ on the pricing page.
Questions · the product
What is it, exactly?
What’s an AI agent, compared to a chatbot?
A chatbot answers when you type at it, then forgets everything when you close the tab. An AI agent has a seat in the office: it sits in meetings, keeps its memory, runs scheduled tasks and reports back. The difference isn’t the model — it’s the seat you give it: the same files and the same rules as the rest of the team. The full picture is on the agents page.
What is Luge in one sentence, for my boss?
Luge is the office where your human team and your AI colleagues work together — same files, same meetings, same rules — with the AI running on your machines or in a private cloud that stays in Canada. Not one more assistant in a tab: an office. Take the tour on The office.
Does it replace my people?
No. Luge takes the repetitive work — transcribing, summarizing, filling, chasing — and leaves the decisions to humans. A human approves the actions that matter, edits the drafts before they go out, and an automation can never answer its own question. A good AI colleague makes your team faster, not smaller.
Questions · trust
Your data
Where does my data go when I talk to the AI?
It depends on the brain you pick. With a local model, inference happens on your machine — essentially nothing leaves it. With your own key (Anthropic, OpenAI, Azure, xAI), your requests go to your provider, under your own agreement. With the managed sovereign modules, they stay on Canadian infrastructure. The full journey is mapped on privacy & security.
How does Luge help with Law 25 / PIPEDA compliance?
By design: local processing where you want it, personal-information detection with three responses to choose from — warn, anonymize or block —, an encrypted audit log with a signed exportable archive, and an inference option hosted in Canada. No software can “guarantee” your compliance — it remains an organizational responsibility — but Luge is built to make your life easier in front of an auditor. The full dossier is on privacy & security.
Can the AI see things I’m not allowed to see?
No. An AI colleague acts with the rights of the person talking to it — never more. Semantic search stops at the same boundaries: it never answers from a document or a note you couldn’t open yourself. Your permissions are its permissions.
Who can read my notes?
You, by default. A note stays private until you explicitly share it, and every version in its history names who wrote it — a person, an AI acting for them, or an access key. Outside guests must prove their address (Google account, Microsoft account or one-time code) and only reach what was explicitly shared with them.
Questions · meetings
Meetings
Does a bot join my calls?
Only if you choose it — and never covertly. Three paths: the meeting happens in a Luge voice or video room, where the note taker is an invited participant anyone can stop; the capture happens locally from your own device, with nothing joining the call; or a visible Teams bot goes through the lobby, where you admit it, refuse it or remove it. The three paths are detailed on the meetings page.
Who said what — is speaker attribution reliable?
In Luge rooms, yes — for real. The system listens to each person separately instead of guessing who’s speaking on a mixed track: per-speaker attribution is real, not estimated. For a capture made from your device on an outside tool, attribution relies on diarization — recognizing voices on the recording: useful, but estimated.
Does transcription work in French?
Yes — as long as you pick the right engine. The default local transcription (Parakeet) is excellent in English, but drifts toward English on weak-signal French. For reliable French, force Canary or Cohere in the settings — it’s a menu, not gymnastics. We’d rather tell you up front than let you find out on meeting minutes.
Questions · the engine
The AI itself
Which model runs underneath?
The one you choose. Plug in your own key — Anthropic, OpenAI, Azure or xAI — and that’s the model doing the work, under your agreement. Or stay local with small and mid-size open-source models, the class that runs on a good workstation — we’ll never pass them off as frontier models. The honest picture is in the local models guide.
Does it learn from my data?
Not by retraining: there’s no fine-tuning on your data. Luge improves through its memory — lessons drawn after each task, attributed to their source, stored on two floors (Core and Pool). The nightly sweep sorts, consolidates and archives — it never deletes anything. You can see where every memory came from.
Can an agent act without my permission?
Not for anything that matters. Sensitive actions go through a human step: the agent drops a task in your inbox, with a draft you edit before approving — not just a yes/no button. And the house rule is non-negotiable: an automation can never answer its own question. A human decides, every time.
Questions · day to day
In practice
How does it install?
Download the signed installer for Mac, Windows or Linux — direct link, no server to set up — and your first agent is working in two minutes. The web app is there too for teams. Everything is on the download page.
What goes where: note, task, table or document?
Pick the surface by how long the information lives. An idea or a decision that must last: a note (versioned, attributed). An action with a deadline: a card on a kanban board. Data that comes back in rows and columns: a table. A reference document your AI colleagues should read: the knowledge base. And a text you’re building together: a canvas.
Can a coding agent work on MY machine?
Yes. You pair your computer, and a coding agent works right on it — on your machine, with your rights, framed by permissions you set in the conversation thread. Let’s be honest: it’s your workstation, not a sealed vault — you decide what the agent can touch. It starts with the download.
Your question isn’t here? Write us: luge@nsolutions.ai. And if it’s about money, the pricing FAQ doesn’t beat around the bush.