onUI

onUI is a free and open source browser extension designed to facilitate UI annotation for AI agent workflows. It enables users to annotate elements and draw regions directly on web interfaces, supporting workflows where feedback and iterative improvements are central, such as UI development and review. The tool provides two main capture modes: Annotate mode for targeting individual elements and attaching intent and severity, and Draw mode for marking broader layout or spacing regions using rectangles or ellipses. These features allow users to deliver precise feedback on both element-level and region-level UI issues without modifying application code or relying on a backend service.

The extension is compatible with Google Chrome, Microsoft Edge, and Mozilla Firefox. Chrome and Edge installations are available through their respective web stores, while Firefox requires a manual installation process. onUI operates locally, with all components—including the extension, native host, and local MCP server—running on the user's machine. This local-first approach ensures that annotations and data remain private and do not leave the device. The tool does not require any cloud dependencies or SaaS integration, and it can be enabled on a per-tab basis, giving users granular control over which pages are annotated.

onUI's workflow is designed to support a continuous loop between user annotation and AI agent action. After annotating, an AI agent can access fresh annotations via local MCP tools such as onui_get_report, apply suggested fixes, and repeat the process until the UI meets quality standards. The extension offers four export formats—compact, standard, detailed, and forensic—catering to different levels of reporting granularity. The interface features a compact toolbar with primary actions and a pop-out card for settings, including output level and clear-on-copy controls. Keyboard shortcuts are available for quick access to annotation modes, and shortcut customization is supported across all browsers.

onUI is positioned as a tool for developers and others involved in UI iteration and automation, particularly where annotation-driven feedback loops with AI agents are valuable. Its open source nature and privacy-focused design make it suitable for local development environments where data control is a priority.