Ready to grow with AI? Let's talk.
Book Free ConsultationA personal AI operating system — an iPhone app that talks to a live "brain" running on a Mac, with voice control, persistent memory and task automation.
Live screens from the current build — Jarvis is still evolving, so expect this gallery to grow.



Jarvis is a personal AI operating system I'm building for myself — one assistant that lives on my iPhone but actually thinks on my Mac. The goal: a single interface for voice commands, reminders and calendar management today, growing into full oversight of my other projects — Aegis AI and Slay Beauty London both already have dedicated "Workspace" modules planned — all wrapped in a live, neural-style dashboard that shows exactly what the system is doing in real time instead of hiding behind a spinner.
Jarvis is split into two halves that talk to each other over the local network — a lightweight phone app and a much more powerful Mac server doing the actual thinking.
The dashboard, voice/text input, memory browser, tools grid and settings all live here. It's built to visually show system state — CPU, memory, network and battery of the connected Mac — plus a live "Idle → Listening → Analyzing → Thinking" status so the assistant never feels like a black box.
A local Python API (server v2.0.0, API v1) runs the actual reasoning using GPT-4o-mini, with the model swappable via a .env config. It holds persistent memory between conversations and exposes a dedicated /voice endpoint for Whisper speech-to-text and text-to-speech.
The phone finds the Mac over the local network via a saved server address and device token, with Reconnect and Test Connection tools built in for when the connection drops — no cloud server required, everything runs on hardware I own.
Running a genuinely capable AI assistant entirely on a phone is either slow (constant cloud round-trips) or weak (on-device models can't reason well). I wanted real reasoning power without giving up mobility — and I wanted to actually see the assistant think, not just receive a black-box reply after a loading spinner.
Jarvis splits the work: the iPhone app is purely the interface, while a Python server on my Mac holds the real brain (GPT-4o-mini today, swappable later) plus a Whisper + TTS voice pipeline. The phone talks to the Mac over the local network, so I get full reasoning power alongside a live "Thought Process" panel instead of a black box.
Jarvis ships in small pieces — here's exactly what's working today and what's coming.
Jarvis is a live, evolving personal project — new tools and wake-word voice mode are shipping incrementally. Want an AI system like this built for your business?
Book a free discovery call and let's talk about your project.