AegisAI

AegisAI is a privacy-first AI assistant that runs entirely on your own machine. Ask questions, summarize any web page, and get answers from a local AI model — with no account, no cloud, and no telemetry. HOW IT WORKS AegisAI runs a compact AI model directly inside your browser using WebGPU (with an automatic CPU fallback). After a one-time setup, everything happens on-device. - First-use setup: On your first request, AegisAI downloads its ~360 MB offline engine from Hugging Face. This is the only time an internet connection is needed. - After that: The model is cached locally and AegisAI works fully offline — your prompts and the AI's responses never leave your computer. WHAT YOU CAN DO - Chat with a local AI model from the toolbar popup - Summarize the page you're reading, with the result shown in an on-page panel - Work completely offline once setup is complete PRIVACY BY DESIGN - No telemetry and no analytics — we collect nothing - No account and no sign-in - Your prompts and conversations are processed locally and are never transmitted to us or any third party - Optional encrypted local storage for anything you choose to save - The only network request AegisAI makes is the one-time model download described above ADVANCED: BRING YOUR OWN OLLAMA Power users can switch to "Advanced Ollama Mode" in settings to run inference against a local Ollama instance instead of the built-in engine. This is entirely optional and off by default. PERMISSIONS - Storage — to cache the model and save your settings locally - Active Tab + Scripting — used only when you click "Summarize this page," so AegisAI can read that page on demand. AegisAI does not run on pages in the background and requests no broad site access. AegisAI is open about what it does: local-first AI, no tracking, and a clear one-time setup. Install it, send your first message, and after a short download it's yours — private and offline.