Running agents locally

Since the servers are always “exhausted,” could it be possible to integrate local Ollama models for both offline and a more stable agent experience, or to just use your own OpenAI, Gemini, Anthropic, etc., API key?