Features
Quenderin isn't a stripped-down toy because it runs locally — it's a capable assistant that happens to need nobody's servers. Here's what that buys you.
After the one-time model download, there are zero network calls. It answers at 35,000 feet, on a train through a tunnel, or three days into a hike with no bars. Offline isn't a degraded mode — it's the normal mode.
No subscriptions, no per-token billing. Inference runs on hardware you already own.
Quenderin includes a tool-using agent that can carry out small tasks — but it ships with a hard safety blocklist it can never override on its own. It will not autonomously pay, delete, transfer money, or touch credentials. Sensitive actions are surfaced for you to confirm, never taken silently.
No accounts, no analytics, no telemetry. Your conversations stay on your device — and because it's open source, that's verifiable, not a promise.
It picks a model that fits your memory and chip — and warns you before one that doesn't, so it never stutters or fails to load.
MIT licensed. Read every line, audit every privacy claim, and build it yourself from source.
Models keep downloading in the background and resume after an interruption — switch apps the night before a trip and it's still ready by morning.
The difference
| Quenderin | Cloud assistant | |
|---|---|---|
| Where it runs | On your device | A remote data center |
| Works offline | Always | Never |
| Your conversations | Stay on your device | Sent to a server |
| Cost | Free — no tokens | Subscription + per-token |
| Account | None | Required |
| Open source | Yes — MIT | Closed |
Every feature above works with any model you choose — from a Raspberry Pi build to an M-series flagship.