Sentient OS: An Open-Source, Proactive Local AI Agent for macOS

A developer has introduced Sentient OS, an open-source, local AI agent that runs on-device to index personal data like emails, messages, and screenshots. It proactively suggests and executes tasks overnight by waking up a Mac at 3 AM to build a localized knowledge base. This project demonstrates a practical use case for local LLMs, addressing the privacy concerns and high API costs associated with cloud-based AI agents. By keeping raw data on-device, it enables a highly personalized assistant that can safely interact with sensitive user information. Sentient OS runs a custom Gemma 4 model on a LiteRT-LM fork for 90% of its workload, relying on a user-provided frontier model (like Qwen 3.7 or Kimi K3) for the remaining 10% of complex reasoning. It requires Apple Silicon (M1 or newer), macOS 15+, and at least 8 GB of RAM.