An MCP server that bridges AI-powered applications with Fibaro HC3 smart home systems, enabling natural language interactions to manage rooms, devices, and scenes. It implements the MCP protocol with actions grouped under Room Management (list_rooms, get_room), Device Control (list_devices, get_device, call_device_action), Scene Management (list_scenes, get_scene, execute_scene, kill_scene), and System Tools (test_connection). The server talks to Fibaro HC3 via its REST API (rooms, devices, scenes) and exposes an MCP transport layer that can run over STDIO or HTTP. When using HTTP transport, the server provides an endpoint at /mcp with POST for client requests, GET for server-to-client SSE notifications, and DELETE to terminate sessions; requests can be authenticated with an x-api-key header. The project includes quick-start steps, configuration via a .env file (HC3_HOST, HC3_USERNAME, HC3_PASSWORD, MCP_TRANSPORT_TYPE, etc.), and Docker deployment options. It also includes integration examples (Claude Desktop) and an MCP Inspector development tool, which exposes a debugging web interface at http://localhost:6274/ during dev. It leverages the Fibaro HC3 REST API paths: /api/rooms, /api/devices, /api/scenes.