INVIZ — Eyes-Free Internet for Everyone is an open-source, AI-powered voice browser assistant built to make the modern web more accessible, operable, and safer for people who cannot rely on vision. Instead of merely reading webpages like a traditional screen reader, INVIZ understands the live page, discovers actionable elements, performs clicks, typing, navigation, searches, and form interactions through voice, and narrates what it is doing. Its ContextLens provides live, accessibility-aware page context, while the agent follows a bounded Reason → Guard → Execute → Verify architecture so that actions are grounded in real page elements rather than hallucinated interactions. Security is a core design principle: WebGuard policy-checks every proposed action, sensitive fields receive additional protection, consequential actions require explicit user confirmation, and execution is bounded by action, reasoning, and time budgets. INVIZ follows a fail-closed philosophy—when the system cannot reliably observe, ground, execute, or verify an action, it should stop rather than fabricate success. Provider credentials are isolated from the browser through a local backend gateway, while local-first AI reasoning through Ollama reduces unnecessary cloud exposure, with controlled cloud fallbacks when required. The system is built as a Chrome MV3 extension with a TypeScript frontend, content-level page understanding and execution, an offscreen voice pipeline, and a lightweight backend connected through versioned APIs. Crucially, INVIZ is designed as a page-agnostic system rather than a collection of website-specific scripts: it discovers the structure, controls, and state of the current webpage and reasons over what it actually observes, allowing the same architecture to work across familiar and unfamiliar websites. As an open-source community project, INVIZ is intended not only as a finished accessibility tool but as a foundation that developers, accessibility researchers, and contributors can inspect, audit, extend, and improve—bringing together AI agents, browser automation, accessibility, security, and human-centered design to move toward a web that can be independently operated by everyone, regardless of whether they can see it.