Local-first macOS agent interface
A local desktop agent you install on your Mac, then command through a live console and voice UI. The website launches the process; the automation stays on your machine.
Shell installer. It clones or updates the GitHub repo, installs Python deps with uv, builds Lumen.app, and places it in ~/Applications.
Installer manifest
The download button fetches a small shell script named lumen-installer.sh. It is not a signed DMG yet, and it is not model weights.
Clones or updates OmTheLast/Lumen into ~/Documents/Code/Lumen.
Runs uv sync, builds Lumen.app, and installs it into ~/Applications.
Ollama, LM Studio, Whisper models, tokens, and private config remain separate local prerequisites.
Use this when the repo already exists on your computer. It starts the Python agent, local web console, and bottom-right desktop orb.
cd ~/Documents/Code/Lumen && uv run python -m lumen.main
After it starts, this page should detect the local console and enable the Open Console button.
Lumen expects local models through Ollama or LM Studio. Model weights are never shipped from this website or stored in the repo.
ollama pull qwen3:latest
Voice support is optional. Install the voice extras only on machines that need local speech transcription.
uv sync --extra voice
Fresh install
For a new user machine, install prerequisites, clone the repo, then start Lumen locally. The public site can guide the install, but the user must approve commands in Terminal.
Download Installer Script/bin/bash -c "$(curl -fsSL https://lumen.ompatnaik.com/install.sh)"
git clone https://github.com/OmTheLast/Lumen.git ~/Documents/Code/Lumen