Right AI

RightAI is a Chrome side panel that puts every AI you use in one place — and adds a real browser agent on top. Two things set RightAI apart: 1. Omni — a browser agent that drives Chrome for you. Give Omni a goal; it reads the page, clicks buttons, fills forms, switches tabs, runs page-scoped scripts, and writes files to an authorized local workspace. It is confirmation-gated: every privileged action waits for your approval in the side panel before it runs. 2. Parallel multi-model chat. Send the same prompt to several models — and any OpenAI-compatible endpoint — in one go. Read the answers side by side and pick the one you trust. Launch RightAI RightAI ships with three commands. The side panel is bound to Ctrl+R. Quick Input to Ctrl+Q (send a prompt with the current page attached). Press Ctrl+Shift+L to soften bright pages for late-night reading. The fourth command — Omni — has no default shortcut because Chrome only allows four suggested keys per extension, and the three above already fill the budget. Open Omni from the side panel toolbar instead. (You can still bind a keystroke manually from chrome://extensions/shortcuts.) Right side panel chat • Multi-turn AI conversations in a clean, resizable side panel. • Page context injected on demand: title, URL, main content, optional web fetch for up-to-date answers. • Works with any OpenAI-compatible endpoint, plus local models, self-hosted servers, and custom bases. If the server exposes models, they appear in the picker automatically. • Attachments: drag in files, paste screenshots, or right-click any image to OCR it inline. Quick Input (Ctrl+Q) Press Ctrl+Q anywhere to send a prompt with the current page attached. Choose between a compact overlay on the page or a dedicated conversation window that keeps your flow open while you switch tabs. Omni — the browser agent Powered by an OpenAI-compatible model. To get started without paying for an API, the Omni settings page includes a step-by-step walkthrough for a free OpenCode Zen key. Omni's toolkit: • Read and write files. read_file, write_file, diff_file. Authorized either as a directory workspace or by selecting individual files. All access is logged. • Drive pages. Screenshot the current tab, run page-scoped JavaScript (cannot touch Chrome APIs or local files), open new tabs, switch tabs, and list tabs. • Compound tasks. perform_task executes multi-step goals end-to-end, with live streaming and progress indicators you can follow in the side panel. Safety Omni is local-first. API keys, settings, and the audit log live on your device. Every privileged action opens a confirmation card in the side panel listing exactly what will run. Shell-level execution, broad page scripting, and background automation are intentionally not supported; you stay in the loop on every step. Built around your workflow • Customizable toolbar. Pin any AI site from the built-in gallery of 30+ apps, or add your own links. Drag to reorder, import and export your layout as JSON. • Flexible layouts. Dock the panel on the right or pin it to the top. Toggle split view when you want two panes, or disable it when you don't. • Liquid-glass UI. Native ES modules, MV3-compliant, theme-aware light/dark, with persistent history and graceful confirmations. • 47 languages. The UI is fully localized, including English, Chinese (Simplified), Japanese, Korean, Spanish, French, German, Arabic, Hindi, Bengali, Portuguese, Russian, and more. Translate, dim, focus • Selection translation. Select text on any page and get an inline bubble translation, or open the side panel to translate with full page context attached. • Image OCR. Right-click any image for a floating OCR overlay with copy-to-clipboard. • Night Dim. Press Ctrl+Shift+L to dim bright pages for late-night reading. • Crash and Jenkins consoles. Optional shortcuts for the times you need them. Privacy and permissions RightAI keeps your work on your device. API keys, settings, history, and the audit log are stored locally via chrome.storage and localStorage; nothing is sent to RightAI's servers (there are none — the extension is fully client-side). Configuration can be exported and re-imported at any time from the settings page. Declared permissions are minimized to what each feature actually needs: sidePanel, activeTab, tabs, storage, scripting, declarativeNetRequest, debugger (page diagnostics only), tabGroups, downloads, and alarms. Content scripts run on `<all_urls>` only because features like selection translation, page context, and Omni need to work on every page you ask about; they are excluded from the Chrome Web Store (`https://chromewebstore.google.com/*`, `https://chrome.google.com/webstore/*`) in `manifest.json#content_scripts.exclude_matches`, and Chrome's content-script runtime never injects them into privileged pages like `chrome://*/*`, `chrome-extension://*/*`, `chrome-search://*/*`, `chrome-devtools://*/*`, or `edge://*/*`. Open the side panel with Ctrl+R and start exploring.