Open source · every token computed in your hand
Quenderin runs open models — Llama, Qwen, Gemma — fully on your device. No cloud, no account, no telemetry, and it works offline. On a Mac it can also act on your behalf: files, apps, Apple Shortcuts, asking before each change. Free, and every line is on GitHub.
Built on
A mainstream assistant is a thin client for a data center: lose the connection and it goes dark; keep it and your words travel to hardware you’ll never see. Quenderin is the model itself, living on the device in your hand. It’s smaller than the cloud ones and it will sometimes be wrong — we publish exactly how much smaller. It is also always there, and it is only yours.
How it works
Quenderin reads your hardware—RAM, chip, GPU—and recommends the largest model your device can run well.
Pull the model once over Wi-Fi—0.4 to 4.7 GB, your choice. It resumes in the background if you get interrupted.
Everything runs locally from then on. A clear check tells you it’s safe before you leave the grid.
The computer-use agent
On a Mac, tell Quenderin a goal and a local model operates the computer for you — organizing files, driving any app, running your Apple Shortcuts. A private alternative to cloud agents like Cowork, governed by a trust loop a datacenter can’t offer:
quenderin history), and undo a whole task — even in a new session (quenderin undo).--dry-run shows exactly what it would do, touching nothing.$ quenderin do "organize my downloads" \
--workspace ~/Downloads
· listed 42 files
● Move "invoice.pdf" into "Finance/"?
Allow? [y/N] y
✓ Moved "invoice.pdf" into "Finance/".
● Move "IMG_2291.HEIC" into "Photos/"?
Allow? [y/N] y
✓ Moved "IMG_2291.HEIC" into "Photos/".
● Rename "final_v3_FINAL.pdf" → "report-2026.pdf"?
Allow? [y/N] y
✓ Renamed → "report-2026.pdf".
Undo everything this task changed? [y/N] n
Later? Run `quenderin undo` to reverse it.Straight from the app
Every image on this page is generated from the shipping SwiftUI code — when the app changes, the site re-renders with it.
Android · Closed beta — open now
The native Android app is in closed testing right now. It takes two steps — Google runs closed betas through an opt-in group, so you join that first.
Join with the same Google account your phone uses — that membership is what makes you a tester. It’s auto-approved, no waiting.
Join the group →Once you’re in the group, accept the test and install. If Play can’t find it yet, give it a few minutes to reach your account.
Become a tester →It’s an early build — expect rough edges, and your feedback shapes what ships. Full walkthrough, and what’s different about the Android port, in the beta guide.
Features
After the one-time download, there are zero network calls. Works at 35,000 feet or three days into a hike.
The model is a file on your disk, and answering is just your own CPU and GPU doing math through llama.cpp — there is no service on the other end. Try it: turn on Airplane Mode and ask anything. The only network request Quenderin ever makes is the one-time model download.
Not a promotion — there is no server to pay for. It runs on hardware you already own.
No API meter, no token quota, no trial that expires — because there is no server farm computing on your behalf. Your machine does the work, and the MIT-licensed code can’t be taken away from you. Free stays free structurally, not promotionally.
A private alternative to cloud computer-use agents: it drives your files, apps, and Apple Shortcuts, asks before every change, and can undo the whole task—while never autonomously touching:
Every change is previewed first and waits for your explicit Allow — dismissing the dialog counts as no. Each action lands in an on-device audit ledger you can read afterwards, and the whole task can be undone in one tap. The blocked verbs aren’t a policy promise: they’re enforced by an on-device safety blocklist that ships in the open source.
No accounts, no analytics, no telemetry. Your conversations are yours.
Conversations are plain files on your own disk — export them, back them up, or delete them like anything else you own. There is no account to breach and no analytics SDK to leak: the app simply contains no tracking code.
It picks a model that fits—and warns you before one that doesn’t.
Quenderin reads your RAM and chip, then tells you honestly which models Fit, which are Tight, and which are Too big — before you download gigabytes. Context size, GPU offload, and thread count tune themselves to the machine, from a Raspberry Pi to an M-series Mac.
MIT licensed. Read every line, audit every claim, build it yourself.
The whole app is public — the inference wiring, the safety blocklist, the consent gates, the audit ledger. Don’t take this page’s word for any claim: read the code and check. github.com/alikatgh/quenderin
The model keeps downloading in the background—even if you switch apps the night before a trip.
A dropped connection resumes from the exact byte where it stopped — quit the app, lose the Wi-Fi, come back tomorrow. And before any model is ever loaded, its SHA-256 checksum is verified, so a corrupted or tampered file is rejected rather than trusted.
Made for
Planes, trains through tunnels, trails, ships, field work — anywhere a signal can’t follow you, your assistant still does.
Legal, medical, journalism, research — work that simply can’t be sent to someone else’s server. Nothing leaves the device.
Students, hobbyists, the curious — a capable assistant with no subscription, no token meter, no account to create.
Built for the edge of the map
The night before, on hotel Wi-Fi, you download a model—and Quenderin keeps the download alive even when you switch apps, then shows a green “ready to go offline”. On the trail, with no bars, you still have a capable assistant in your pocket. That’s the whole point: an AI that doesn’t need the world to be online.
Privacy
Model catalog
Quenderin runs on llama.cpp—so it runs any GGUF model: Llama, Qwen, DeepSeek, Mistral, Gemma, Phi. Here’s the curated shortlist it recommends from, sized for everything from a Raspberry Pi to an M-series Mac.
| Model | Good for | Download | Min RAM | Quant |
|---|---|---|---|---|
| Qwen3 4BRecommended | General-purpose, Apache 2.0 — the current go-to | 2.4 GB | 4 GB | Q4_K_M |
| DeepSeek-R1 7B | Step-by-step reasoning & math | 4.7 GB | 8 GB | Q4_K_M |
| Qwen2.5 Coder 7B | Code generation & tool use | 4.7 GB | 8 GB | Q4_K_M |
| Gemma 3 4B | Multilingual, 140+ languages | 2.5 GB | 4 GB | Q4_K_M |
| Phi-4 mini 3.8B | Efficient, runs well on CPU | 2.3 GB | 4 GB | Q4_K_M |
| Mistral 7B | Fast, capable all-rounder | 4.1 GB | 6 GB | Q4_K_M |
| Llama 3.2 1B | Ultra-light — runs on a Pi | 0.8 GB | 1.5 GB | Q4_K_M |
| Qwen3 14B | Best quality for a strong device | 9.0 GB | 12 GB | Q4_K_M |
+ thousands more from Hugging Face—any GGUF works. Choosing is optional: Quenderin picks the best fit for your device automatically.
FAQ
It mostly comes down to one idea: the model runs on your device, so nothing leaves it. Here’s the rest.
Still have a question?
Ask on GitHubYes. Quenderin is MIT-licensed open source, and inference runs on your own hardware—so there are no token costs and no subscription.
No. After the one-time model download, Quenderin makes no network calls. There are no analytics, no accounts, and no telemetry. Because it’s open source, you can verify this yourself.
The iPhone app is on the App Store, and the desktop app runs on macOS and Linux—all today. The native Android app is in closed beta right now, and you can join the test to run it. It’s the same hardware-detection and model-selection engine, rebuilt natively for each platform.
No. Quenderin scales from a Raspberry Pi to an M-series Mac. It detects what your device can handle and recommends a model that actually runs—down to a 0.4 GB ultra-light build.
The iPhone app is on the App Store today, and the native Android app is in closed beta on Google Play—join the testers group to run it now. The desktop app is open source and runnable from GitHub on macOS and Linux.
Run it today
It’s open source and runs from GitHub right now — on macOS or Linux.
git clone https://github.com/alikatgh/quenderin
cd quenderin
npm install
npm run electron:deviOS on the App Store · macOS & Linux today · Android in closed beta
It’s free on the App Store for iPhone, and open source you can run from GitHub today — desktop on macOS and Linux, with the native Android app now in closed beta you can join.