Show HN: OOMU, a Free Native macOS AI Harness for Local and Cloud Models

Summary

OOMU is a free native macOS desktop AI harness built with Rust and AppKit. It is designed for work beyond chat, including interacting with system applications, operating in the terminal, and compiling Word, Excel, PowerPoint, and PDF documents, while also supporting multi-turn tasks on the Mac. The app has no subscription or paid tier and can run entirely with local models; Gemma 4 E2B QAT is the default model required during setup and can be downloaded automatically. Users can instead bring their own API subscriptions for hybrid workflows, with the author reporting tests with OpenAI, Gemini, DeepSeek, Grok, and Mistral, directly or through aggregators. An optional intelligent routing feature uses a small local routing model to judge prompt complexity and choose local or cloud handling on a turn-by-turn basis. OOMU is described as having zero telemetry, an encrypted local database, no background web services, and an air-gap mode that disables network access. The beta release is 0.2.16, with 12 interface languages, a direct download, and a Homebrew installation option.