Inspector: From visual edits to AI-ready prompts

Edit real pages like a design tool and convert visual changes into structured prompts for AI-assisted iteration.

From visual edits to AI-ready prompts. Inspector turns real web pages into an editable canvas for AI-driven iteration. Instead of describing UI changes from scratch, you can click into a live page, make visual edits with immediate feedback, and convert those changes into structured prompts that are ready for modern AI workflows. What makes Inspector different: - Edit real pages directly, with a workflow that feels closer to design tooling than manual prompt writing - See instant visual feedback while adjusting layout, spacing, sizing, appearance, and content - Capture edits as structured sessions instead of losing context in ad-hoc notes - Export prompts based on actual changes made on the page, not vague descriptions after the fact - Keep a history of recorded sessions so you can revisit, compare, and reuse previous iterations Core workflow: 1. Start a recording session on any page you want to inspect or refine 2. Select one or multiple elements directly on the page 3. Adjust position, size, appearance, spacing, and instructions with live visual feedback 4. Save the session and generate a structured prompt from the recorded edits 5. Reuse that prompt in your preferred AI workflow for design, frontend, or product iteration Built for AI-assisted product work: - Product designers who want a faster path from visual intent to AI-ready instructions - Frontend engineers who need concrete, inspectable change requests based on real UI states - Prompt-driven builders who want structured context instead of manually writing every detail - Teams exploring a tighter loop between browser-based editing and AI-assisted implementation Advanced capabilities: - Multi-element selection and alignment controls for more realistic page editing - Session-based recording model for preserving change history - Side panel editing for focused property updates - Prompt export based on live page mutations - Local Bridge integration for sending generated prompts into supported CLI workflows such as Codex or Claude Code Why the permissions are needed: - Access to page content and structure is required so Inspector can let you select elements and apply visual edits on real websites - Scripting permission is used to inject the editing and recording experience into the current page - Storage is used to save sessions, preferences, and local state - Identity is used for Google sign-in where enabled Inspector is designed for people who want to go from “I know how this page should change” to “I have a structured prompt and a reusable iteration record” without switching between screenshots, hand-written specs, and disconnected tools.