Run the full assistant on Windows, and take it with you on Android. Same brain, same memory, same tools — everywhere you work.
The full native desktop experience — MSIX-packaged Electron app with voice, tray supervision, and the Command Center daily briefing.
Sandboxed via the Microsoft Store
Talk to Jarvis and drive your desktop from your phone. Currently in beta: join the Google Play test programme, then install from Google Play.
Play is a closed test — tap "Become a tester", then install
The Android app reaches your desktop over a private Tailscale mesh — no ports exposed, no data through our servers. Install Tailscale on both devices, sign in with the same account, and Jarvis links up automatically.
Also available as a rich CLI, web dashboard, and voice interface — see the setup guide.
A beginner-friendly walkthrough of the first launch, signing in, and giving Jarvis a brain. No developer setup — the Store app configures itself.
Launch Jarvis from the Start menu. A secure-boot screen appears while the local AI core and dashboard warm up in the background.
Jarvis verifies its local runtime — you see a progress screen.
The AI engine and dashboard start up in the background.
Progress reaches 100% and the command center loads.
The main window opens to sign-in. ~10–15s the first time.
First launch is the slow one
The first start takes about 10–15 seconds while dependencies are verified and cached. Every launch after that is much faster because the check is skipped. A small Jarvis icon also appears in your system tray — the app keeps running there so it's always a keystroke away.
On first run Jarvis opens a sign-in page. Creating an account secures your assistant — your data and conversations stay on your own machine.
Your account is local-first
Jarvis is a single-user product — one private instance on your machine, not a shared cloud service. Signing in protects access to your assistant; it does not upload your conversations anywhere. Your history lives in a local database on your PC.
Jarvis needs one AI model to think with. Pick a free local model with Ollama, sign in to Ollama Cloud, or paste a free cloud key — the setup dialog tests the connection before saving.
Free · fully local · no cloud, private
Ollama runs open models directly on your PC, so Jarvis can work completely offline with zero data leaving your machine. It is the recommended way to start — free, private, and no provider account needed.
Back in Jarvis, open the model setup dialog and choose Use local Ollama. Jarvis detects the running Ollama service and the models you have pulled, then confirms the connection. That's it — you now have a private AI running locally.
Groq has a generous free tier and is extremely fast. Great if you don't want to download a model.
In Jarvis: Connect Groq free key — it validates the key live before saving.
OpenRouter gives you 200+ models behind one key, including free options to start with.
In Jarvis: Connect OpenRouter free key — guided, tested before it's saved.
Not sure? Let Jarvis decide
In the setup dialog, choose Use Jarvis recommended model and it auto-detects the best available option for your machine. You can add premium keys (Claude, GPT, Gemini) any time later from Settings → AI Model — Jarvis routes each query to whichever model fits best.
You're set up. Jarvis is now one keystroke away, everywhere on your desktop.
Use the desktop app, dashboard, CLI, or voice mode. Type or speak your request.
Enable voice in Settings → Voice, then just talk. Fast, sub-second spoken replies.
Chat, Command Center, daily briefing, notes and settings — all built in.
Talk to Jarvis in plain language — no commands to memorize. Want the full command reference and power workflows? Read the Usage Guide.
The common first-run questions, and the quick fix for each.
The very first launch verifies and caches dependencies, which can take 10–15 seconds. If it stays on the boot screen much longer, close Jarvis from the tray and reopen it — the cached checks make the second launch fast.
Jarvis needs one model to think with. Open the model setup dialog and either install Ollama (free, local) or paste a free Groq/OpenRouter key. See step 3 above.
Enable the microphone for Jarvis in Windows Settings → Privacy → Microphone, then turn on voice in Jarvis Settings → Voice. The mic permission is required before hands-free mode works.
Install Ollama, pick it as your model, and set privacy to STRICT. In STRICT mode Jarvis never sends anything to a cloud model — it runs 100% on your machine.
The Microsoft Store updates Jarvis automatically in the background. You are always on the latest version with nothing to do.
Quick answers to what people ask before installing Jarvis.
No. The Microsoft Store version is fully self-contained — it bundles its own runtime and runs the dashboard on Electron's embedded Node. A clean Windows 11 machine just works with no manual setup. Docker is optional and only used if you want the Postgres/Redis containers.
The installer is available from the Microsoft Store at no charge, but a limited-time $9.99 one-time lifetime purchase is required to unlock the full assistant, with no recurring app subscription. AI-model usage is separate: bring your own cloud key and pay that provider directly, or run a local Ollama model with no per-token provider bill. Jarvis AI Assistant adds no markup.
On first launch Jarvis AI Assistant verifies its local runtime and starts the AI core plus the dashboard in the background, which takes roughly 10 to 15 seconds. The secure-boot screen shows progress. Subsequent launches are much faster because the checks are cached.
No. Install Ollama and Jarvis AI Assistant can run 100% locally with no cloud, no data sharing, and no account with a model provider. Cloud models (Groq, OpenRouter, Claude, GPT, Gemini) are optional and only used if you add a key.
The fastest free start is a Groq key (free tier, very fast) or a local Ollama model if you prefer full privacy. Jarvis AI Assistant can also auto-detect and recommend a model for you from the setup dialog.
Only if you choose a cloud model, and even then GUARDED mode redacts secrets before anything is sent. In STRICT mode with Ollama, nothing ever leaves your device. Every external call is written to an audit log you can inspect.
