Media
1 item
Description
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 Rashinban for AI is a Chrome extension that helps you build AI-powered browser automation without relying on screenshots or full DOM dumps. ■What it does If you're building an AI agent with MCP, one of the biggest challenges is giving the model enough information to interact with web pages efficiently. Instead of sending screenshots or entire HTML documents, Rashinban extracts only the information needed for interaction—CSS selectors, coordinates, labels, and element states—and makes it available through a local MCP server. ■Why use it? Passing screenshots or full DOM trees increases token usage and slows down AI responses. Rashinban provides a lightweight representation of the page, allowing your AI to: Locate buttons and form fields Understand which elements are clickable or disabled Interact with web pages using structured JSON instead of screenshots This reduces token costs, improves response speed, and simplifies browser automation. ■How it works 1,Install the Chrome extension. 2,Start the local MCP server. 3,Your AI client calls get_ui_elements. 4,The server returns structured UI information as JSON. ■Compatible with Works with LM Studio, Ollama, and any AI client that supports MCP. ■Privacy Everything runs locally. The extension communicates only with ws://127.0.0.1:3765 (localhost). No external servers, analytics, or tracking are used. ■Languages English, Japanese, Simplified Chinese, Spanish, and French.