Refraktor Pro

Voice & model-agnostic WebMCP agent — managed Claude, GPT & Gemini. A Ziola project.

Refraktor Pro turns any WebMCP-enabled web page into something you can talk to. It discovers the AI "tools" a page exposes and drives them for you, by voice or text, using managed Claude, GPT, and Gemini. No API keys to wrangle, no setup: your subscription includes the models. WHAT IT DOES • Discovers a page's WebMCP tools automatically and lists them in a side panel. • Understands what you ask (typed or spoken) and calls the right tools, chaining multiple steps to complete a task. • Reads results back to you, hands-free. • Lets you switch between Claude, GPT-4o, and Gemini per task - all managed, nothing to configure. VOICE-FIRST Tap the mic, say what you tool or action you want, confirm the transcript, and Refraktor Pro then speaks the answer aloud. Choose your preferred system voice. MODEL-AGNOSTIC One agent, three leading models. Pick the one that fits the task; Refraktor Pro translates the page's tools and your conversation for each provider behind the scenes. WORKS ON ANY WEBMCP PAGE Refraktor is general-purpose - it reads whatever tools a page registers at runtime. Nothing is hard-coded to a single site. PRIVACY BY DESIGN Your prompts, the page's tool data, tool results, and voice transcripts are never stored on our servers. Aggregate telemetry is limited to which model was used, how many tools were called, and success/error tied to a hashed license key, never to your content. Voice input uses your browser's built-in speech recognition. REQUIREMENTS • Google Chrome 146+ with WebMCP enabled. • A web page that exposes WebMCP tools. • A Refraktor Pro subscription ($29/month), managed at refraktor.tech. FREE vs PRO The original Refraktor (free, bring-your-own Gemini API key) is a separate extension. Refraktor Pro adds voice, managed multi-model access, and zero setup. Refraktor Pro is a Ziola project. Manage or cancel your subscription anytime at refraktor.tech.