...A built-in AI assistant reads your live terminal, understands the device you are connected to, and can diagnose problems, explain output, or plan configuration changes - always behind a risk-gate that asks before anything mutating runs. It works with Claude, OpenAI, Gemini, Qwen, GLM and GitHub Models, or with a fully local model via Ollama, LM Studio or vLLM for air-gapped environments. It is extensible with MCP servers and reusable Skills, and it is entirely optional.