Sentient-OS-Labs/sentient-os

An on-device LLM understands your entire life, then proactively offers to get your work done through computer use.

What it solves

Sentient OS addresses the privacy and context gaps in current AI assistants. Most AIs lack deep personal context because users cannot upload their entire digital lives to a cloud chat box, and they are typically reactive rather than proactive. Sentient OS creates a private, on-device memory of a user's life to enable an AI that understands personal context and can proactively perform tasks.

How it works

The system uses a hybrid compute model where 90% of the work happens locally on Apple Silicon Macs.

  1. Local Processing: Every night at 3 AM, a local LLM (Gemma 4 E4B) reads files, screenshots, and messages (WhatsApp, iMessage, Apple Notes) directly from disk. It triages data into "keep," "junk," or "sensitive," dropping the latter two immediately.
  2. Knowledge Base Creation: PII-stripped summaries are sent to a frontier model (via the user's own ChatGPT/OpenRouter/LM Studio subscription) to be distilled into a folder of plain markdown files on the Mac.
  3. Proactive Intelligence: The frontier model analyzes the knowledge base to draft "morning cards"—proactive suggestions and drafted replies based on the user's life.
  4. Sidekick: A voice/text interface that uses "computer use" capabilities to execute tasks within the user's own apps and browser, grounded in the local knowledge base.
  5. Optional Sync: An optional, zero-access encrypted MCP server can mirror the knowledge base to the cloud so other AIs (like Claude or ChatGPT mobile) can access the personal context.

Who it’s for

Mac users with Apple Silicon (M1 or newer) running macOS 15+ who want a highly personalized, proactive AI agent that maintains strict local privacy.

Highlights

  • Privacy-First Architecture: Raw data never leaves the Mac; only PII-stripped summaries are sent to frontier models.
  • On-Device Inference: Optimized for 8GB RAM Macs using a custom LiteRT-LM fork of Gemma 4 E4B.
  • Proactive Agency: Automatically drafts tasks and replies overnight so they are ready by morning.
  • Computer Use: The Sidekick agent can interact with Mac applications and browsers to complete tasks.
  • Transparent Memory: The knowledge base is stored as an editable folder of markdown files, not a black box.

Related

  • Project
    osaurus-ai/osaurus

    A native macOS AI harness that provides local agents, memory, and tool execution in a secure sandbox, allowing users to maintain their AI identity and context privately on-device.

  • Project
    johnbean393/Sidekick

    A local-first AI assistant for macOS that provides private, RAG-enabled chat with local LLMs and integrated agentic research tools.

  • Project
    google-ai-edge/gallery

    A mobile and macOS application for running open-source LLMs locally on-device, providing a private and offline environment to experience and evaluate generative AI.

  • Project
    Beingpax/VoiceInk

    A native macOS application that uses local AI models to provide instant, privacy-focused voice-to-text transcription and a conversational AI assistant.