angiejones/mcp-selenium ? reverse-engineered prompt

Build me an MCP server that lets AI agents control Selenium in a simple way, so I can open a browser, go to a site, click around, type into fields, upload files, handle alerts, switch tabs and frames, take screenshots, and even run a bit of JavaScript when needed.

I want it to work with Chrome, Firefox, Edge, and Safari, and to feel easy to connect from MCP clients like Claude Code, Cursor, or Goose. Please include the basic browser session management too, like starting and closing a session, plus a few handy read only resources for the current browser status and page accessibility info.

Make it usable from the command line with a package install flow that just works, and if you need to check the latest docs for Selenium or MCP details, go ahead and look them up online. Keep it clean, practical, and focused on browser automation for AI agents.