Run a full AI agent on mobile with local LLM calls, on-device memory, code execution, and native HTTP support in Capacitor apps.