What problem does it solve? Screenshot-based browser automation consumes large amounts of tokens and produces brittle interactions. This Skill teaches AI agents to use WebMCP with Chrome DevTools MCP so websites expose structured JavaScript functions as AI-callable tools, reducing token usage by up to 89% compared to screenshot workflows. ## Core Features & Use Cases - WebMCP Tool Registration: Register JavaScript functions as AI-callable tools via navigator.modelContext with JSON Schema inputs, error handling, and structured responses. - Tool Discovery and Execution: Use the Chrome DevTools MCP tools list_webmcp_tools and call_webmcp_tool to discover and invoke registered page tools from Claude Code, Cursor, Claude Desktop, or Windsurf. - Browser Automation: Navigate pages, click elements, fill inputs, take screenshots, and evaluate scripts through 26 Chrome DevTools MCP tools connected over the Chrome DevTools Protocol. - Use Case: Build a calendar web app where an AI agent discovers a create_calendar_event tool, calls it with title, date, and time arguments, and verifies the result, all without parsing screenshots. ## Quick Start Set up WebMCP with Chrome DevTools MCP in my project and register a JavaScript function as an AI-callable tool.