Chromex

由 Codex 驅動的瀏覽器助理，可處理頁面脈絡、分頁、語音與圖片編輯。

Chromex brings a Codex-powered assistant into Chrome’s side panel so you can work with the page you are already viewing. Ask Chromex to summarize webpages, compare selected tabs, explain page content, draft replies, turn YouTube videos into timestamped notes, attach files, dictate prompts, and edit or generate images from page context or uploaded references. Key features: • Page-aware chat: ask questions about the current page, selected tabs, uploaded files, screenshots, and images without switching apps. • Hybrid page understanding: Chromex can use page text, selected DOM content, visible-screen context, or supported site adapters depending on the task. • YouTube workflows: summarize videos, explain the current moment, extract chapter notes, and jump to timestamps when available. • Image workflows: edit images from the page, upload reference images, create infographics, and open generated assets from the local output folder. • Voice and dictation: dictate into the chat box or start a live voice session when your local Codex runtime supports it. • Profiles and skills: choose reusable instruction profiles for research, marketing, product work, support, fact-checking, slide planning, and more. • Local bridge architecture: authentication, Codex app-server communication, temporary files, generated images, and diagnostics are handled by the local native bridge rather than by Chrome extension storage. Chromex is designed for user-controlled workflows. Browser history, microphone access, host permissions, tab context, screenshots, and page actions are requested only for the feature being used. The extension does not sell data, does not run ad tracking, and does not store raw OpenAI API keys in Chrome extension storage. Requirements: • Chrome 116 or later. • The Chromex local native bridge installed on the same computer. • A supported Codex/OpenAI account or API-key fallback configured through the local bridge. Install Chromex if you want a practical browser assistant that can understand the page in front of you, help with real web workflows, and keep sensitive runtime logic behind a local bridge.