Open source · runs on your machine

My Jarvis

Just A Rather Very Intelligent System

A voice-first AI assistant that sees your screen, remembers what matters and controls your system — powered by the AI provider you choose.

1command
6AI providers
20features
MITlicense

Not a chat window

It seesReads your screen on request or continuously.
It remembersFacts you tell it, stored encrypted on your disk.
It actsOpens apps, types, clicks — and works through your to-dos.
What it can do

Twenty modules, all optional

Skip what you don't set up — My Jarvis starts anyway.

Voice & chat

Talk or type, in six languages. Spoken replies optional.

Screen vision

Screenshot analysis, or live vision every 2–3 seconds.

Code that runs

Every snippet is syntax-checked, fixed and formatted.

Calendar & tasks

Google Calendar, Todoist and Notion in plain language.

Smart home

Home Assistant and Philips Hue: lights, heating, media.

Kill switch

Ctrl+Alt+J stops everything, instantly.

Safety

You keep control

Kill switch

One shortcut stops every running action.

Risk checks

Dangerous commands are blocked, risky ones ask first. Emails are never sent unconfirmed.

Your disk

Config, memory and keys live in ~/.jarvis/, encrypted.

Your prompts — screenshots included — go to the provider you pick. Run a local Ollama model to keep everything on your machine.
Help

Common snags

“API key missing” banner

Open settings with Ctrl + ,, pick your provider, paste the key.

Microphone not working

Install PyAudio — on Windows via pipwin if the direct install fails. Without it My Jarvis falls back to sounddevice.

Terminal
pip install pyaudio

# Windows fallback
pip install pipwin && pipwin install pyaudio
No spoken replies

Check that TTS is on in settings. On Linux, install the speech engine:

Linux
sudo apt-get install espeak-ng
Port 8765 is taken

My Jarvis tries 8766, 8767 and so on by itself. If it still fails, close whatever holds those ports.

Wake word doesn't respond

Enable it under Settings → Activation. Without pvporcupine it uses Google Speech Recognition and needs internet.

Talk to your computer