Rashinban for AI

AIエージェント向け：ページ内の操作可能要素のCSSセレクタ・座標・状態を抽出

Rashinban for AI is a Chrome extension that helps AI agents interact with web pages efficiently — without screenshots, without full DOM dumps. What it does When an AI agent needs to click a button, fill a form, or navigate a page, it typically relies on screenshots or full HTML parsing. Rashinban takes a different approach: it extracts only the interactable elements (CSS selectors, coordinates, and states) and delivers them via a local MCP server. Why it's faster and cheaper Screenshots and full DOM transfers consume significant tokens and processing time. By passing only what the AI needs to act — selectors and coordinates — Rashinban reduces token usage and speeds up response time, making AI-driven browser automation more practical for everyday use. How it works Chrome extension monitors the page for interactable elements A local MCP server receives the extracted data via WebSocket Your AI agent calls the get_ui_elements tool and gets structured JSON back Compatible with LM Studio, ollama, and any MCP-compatible AI client. Privacy All communication stays on your machine. Data is sent only to ws://127.0.0.1:3765 (localhost). No external servers, no analytics, no tracking. Languages Available in English, Japanese, Chinese (Simplified), Spanish, and French.