A role
- ops assistant
- analyst
- researcher
You hire it for a specific job, not for small talk.
Summit · the start
Hire your first AI employee.
It works on your machine — not in someone else's cloud. Meetings recorded with no bot on the call. AI that runs locally. A price that never moves.
Luge, a product of N Solutions. Built in Québec. Already in production at its first vertical customer: Tchat N Sign (financial services) — an organization certified SOC 2 Type II, ISO 27001 and ISO 27701.
In short
Luge is an AI employee that works on your machine: meetings recorded with no bot joining the call, open-source models running locally or on managed Canadian infrastructure, team memory that improves with use without retraining — at a flat price, never per token. Free on Solo with your own AI; built in Québec by N Solutions.
Turn 1 · the idea
A chatbot answers.
An employee gets things done.
Luge isn't a tool you invoke — it's a colleague you onboard. You don't prompt it, you give it a job: like an intern who never forgets, never sleeps, and actually reads every document you hand over.
A memory
- team memory
- learns from every task
It remembers last week — and gets better each run.
Autonomy
- scheduled tasks
- delegation
- escalation
Give it a goal, not a script. It escalates when it should.
And it improves on its own. No retraining.
After every task, Luge extracts what it learned and keeps it. No fine-tuning, no magic: well-architected memory. The employee you have in January isn't the one you have in June.
- It does the job meeting, report, email triage — whatever it is
- It extracts a lesson automatically, after every task
- The lesson joins team memory attributed to its author, searchable by every agent
- The next task starts with the experience yours, and the whole team’s
A deliberately small notebook
The essentials live in a notebook with a hard budget. When it's full, something has to go — that's what keeps it reliable. Try it:
Core memory — the little notebook
1110 / 1200 characters
- Always answer Beauce clients in French. 420
- Reports go out Monday 7am, PDF format. 360
- Reports are sent Monday morning as PDFs. 330
Every memory says who taught it
Team knowledge is cited like a proper report — human and AI, never conflated.
“Client Tremblay always wants his documents as PDFs.”
“The weekly report takes 4 minutes on the local model.”
A future reader will never see an AI statement quoted as a human one. Because you stay accountable for your advice — not your AI.
While you sleep
At night, it tidies up.
Every night, its memory reorganizes itself: duplicates merge, contradictions get resolved, and a personal memory that leaked into the shared pool returns to your private drawer on its own. It reorganizes — it never invents. The curious can read how it actually works.
Team memory
“Client Tremblay prefers PDFs”
“Willie leaves early on Thursdays”
Private drawer
Turn 2 · the install
Installed in 2 minutes.
No IT department cosplay.
No terminal, no VPS, no tokens to paste into config files. A signed installer, a two-click device pairing in your browser, and a guided onboarding in the app. We love DIY agent projects — we compare a few of them here, respectfully — but your evening is yours.
- 1 Download the signed installer
- 2 Approve the device in your browser
- 3 Your first AI employee clocks in
Turn 3 · meetings
No bot ever joins your call.
Everyone has seen "a bot has joined the meeting" and felt the room stiffen. Luge detects Meet, Teams, Zoom, Webex or Jitsi and records locally on your machine — nothing joins the call, nothing shows in the participant list. Then: transcript, summary, action items. Details here.
Recording a conversation is still your responsibility: tell the participants. Your device, your rules, your jurisdiction.
Night · your data sleeps here
The AI runs on your computer.
Like software used to.
Small, fast open models (Qwen, Ornith — GGUF via llama.cpp) handle the office work: not frontier magic, reliable work. Transcription happens on-device. Your secrets stay in 1Password. Your files, your audio, your keys: processed on your device. The curious get a local models guide, skeptics can go read under the hood.
- Your documents
- Your meeting audio
- Transcription (Parakeet)
- Local models (GGUF)
- Your secrets (1Password)
Processed on your device. Full stop.
- Calls to YOUR API key (Claude, OpenAI…) — under your agreement
- Team sync, if you enable a shared workspace
- Nothing at all, if you run 100% local
Always your choice, never a hidden default.
Turn 4 · the bill
A bill that never surprises you.
Per-token billing is a taxi meter: one busy month and the bill explodes. Luge inverts it: small models, orchestrated into workflows where every step is short, scoped and verifiable. The result — predictable costs, and an AI that hallucinates a lot less, because it never improvises in one giant blob.
±42 % month-to-month swing, billed per token
0 % surprise — small models, scoped workflows, flat price
Turn 5 · your AI
Bring your own AI.
Like a bring-your-own-wine restaurant.
We provide the table, the service and the menu — the bottle is yours. Plug in your Anthropic, OpenAI or Azure key, point at your vLLM server, or download a local model right in the app. No key at all? Run 100% local. That's the Solo plan: free, because your machine does the work.
- Anthropic Claude
- OpenAI
- Azure OpenAI
- vLLM
- Local GGUF models
Turn 6 · the day job
One employee, forty-three tools.
Email, calendars, tickets, CRM, code repos: your AI employee plugs in where the team already works, runs its scheduled tasks, reads your documentation (hand it the binder — it reads the whole binder), and codes in an isolated sandbox.
- scheduled task · monday 7am — "weekly quotes report"
- read: Google Drive /Quotes (12 files)
- analyze: local model qwen-8b · 100% on-device
- report generated → posted to Slack #ops
- call_colleague: phoned Marc — meeting confirmed
- memory: "client Tremblay prefers PDFs" — noted
- done in 4 min · marginal cost: $0
21 native integrations + 22 platform modules via MCP — plus an OpenAI-compatible API to wire up the rest.
Last turn · what's next
Growing? Add colleagues.
Regulated? Keep it in Canada.
Finish line
Come see what shipped this week.
The product moves fast — the changelog is public, the distribution repo is on GitHub, and the orchestration foundation, RoomKit, is open source.