Skip to main content
Lethean lthn .ai
Sign in Open console
lthn.ai Desktop
macOS tray · local LLM · single-watt

Desktop

lthn is a native macOS tray app that runs local LLM inference on your own hardware — sovereign compute, single-watt, airplane-mode capable. The tray is the app; chat, benchmark, and telemetry open as windows from it.

On-device · Apple Silicon No cloud round-trip EUPL-1.2
WHAT IT DOES

Sovereign compute, on your hardware

Local inference

A go-mlx runner serves models on Apple Silicon — start, stop, and generate on your own hardware, with no cloud round-trip.

Single-watt telemetry

Live watts and memory from powermetrics and IOReport, so the energy cost of every token is on screen.

Tray-rooted

A menubar icon and a 400×560 popover. Chat, benchmark, and settings open as windows from the tray; closing them does not quit the app.

Talks to the platform

The bridge connects the app to api.lthn for plans, sessions, the agent fleet, and the marketplace.

WINDOWS

One symbol, six routing surfaces

Everything opens from the tray as a transient window, anchored to the running process.

Chat Model browser Benchmark Telemetry Fine-tune Agent fleet Tools Network
BUILT WITH

Built with

Go + Wails 3
Tray and windows
Lit
Web Components
go-mlx
Metal inference
api.lthn
Platform bridge
In Development

Builds soon

The tray, runner, and telemetry are coming together. Signed macOS builds will land at download.lthn.ai when the first release is cut.

Follow on GitHub