"author:Garnier4743" matching MCP servers:

Enables an LLM to automate a real browser by describing one outcome per step, with page content summarized into compact element lists and a fast Typesafe System One model deciding the target element, action, value, and whether the step is done, blocked, or needs confirmation, while Playwright performs the action. Also supports direct element actions, yes/no and choice questions, relevance-filtered page reading, screenshots, and calling tools that a WebMCP-enabled page registers.