Jarvis AI Assistant gets work done on your Windows PC.

Your conversations, memory and settings are stored on your PC. What reaches a cloud AI model depends on the privacy mode you choose in Settings. RELAXED, the default, sends prompts to the cloud provider you picked with no redaction. GUARDED strips credential-like strings such as API keys, passwords and tokens first; email addresses and phone numbers are not redacted. STRICT uses only local Ollama models, though online tools such as web search still work. A local log records which provider and model were called, not the content.

Anthropic (Claude), OpenAI (GPT), Google (Gemini), Groq, OpenRouter for 200+ more, and local models through Ollama or the built-in llama.cpp engine — all through one unified, prefix-based router. Switch models with zero config.

Jarvis AI Assistant ships as a native Windows desktop app (Microsoft Store / MSIX), plus a CLI, web dashboard, and voice interface. Apps for macOS and Linux are coming soon.

Jarvis AI Assistant is an agentic system, not a chat window. It runs 70+ real tools, remembers across sessions, routes between 7+ AI providers, and runs on your own hardware with privacy controls. It does work — it does not just talk about it.

Only so it can do the calendar tasks you ask for, and connecting is optional — Jarvis works fully without it. Calendar access lets it read your schedule, create, update, and delete events, and find free time. It does not request any Gmail permission. Reading, searching and sending email is a separate, optional step that uses IMAP and SMTP with an app password you create yourself and can revoke at any time — which also means it works with Gmail, Outlook, iCloud, Fastmail and any other provider. Jarvis AI Assistant only accesses Google data after you explicitly grant permission on Google's own consent screen, and you can disconnect at any time from inside the app or by revoking access in your Google Account permissions.

Partly. A local log records which cloud provider and model were called and how many messages were sent, but not the content of those messages. Your conversations, memory, settings and logs are stored in the .jarvis folder in your Windows user profile, where you can see them.

Yes. Free-Hand voice mode does wake-word → live conversation → proactive idle, with fast sub-second first audio using ElevenLabs (premium, opt-in) or local Faster-Whisper + Kokoro.

No. Talk to it in plain language from the desktop app or voice. Developers get a powerful CLI and tool registry; everyone else gets a Command Center with daily briefings, notes, and reminders.

Yes — they are called routines. You describe the steps in plain language ("check my unread email", "send me a short summary") and say when ("every morning", "every day at 5pm", "every 2 hours"). Jarvis then runs those steps on its own, using the same tools and the same model it uses when you chat with it. You can pause, resume, run it now, reschedule, or delete a routine at any time.

It catches up once. When Jarvis starts again it detects the missed schedule and runs the routine a single time, rather than firing every run you missed in a burst. Routine schedules also resolve against your machine’s local timezone, not UTC, so "9am" always means 9am where you are.

No. PostgreSQL is optional everywhere in Jarvis AI Assistant. If you have not connected one, routines and their run history persist to a local SQLite file in your home folder. There is nothing to install and nothing to configure.