The browser-use-mcp-server is an MCP server that enables AI agents to control web browsers through the browser-use framework. It leverages Playwright for browser automation and can operate in two transport modes: SSE (server-sent events) and stdio. The project supports real-time VNC streaming so users can observe browser automation as it happens. Prerequisites include uv (a Python package manager), Playwright, and mcp-proxy for stdio mode. Environment configuration is via a .env file, with OPENAI_API_KEY required and optional CHROME_PATH and PATIENT flags (PATIENT=false by default) to determine whether API calls should wait for task completion. Installation involves installing dependencies with uv, installing Playwright browsers, and optional docker usage. The server can be run in SSE mode (uv run server --port 8000) or in stdio mode (build, install the wheel, then run with --stdio and a proxy port). A Docker section provides image-based deployment with default VNC password ("browser-use") and an option to mount a custom password secret. The README also covers a VNC viewer setup, sample client configurations, and local development steps for building, testing, and updating the server.