Open models can run inside Claude Desktop, with Ollama acting as a third-party gateway provider for the app. Setup happens on Ollama's side: open Ollama, select Claude, switch it on, and the gateway gets configured for you.
While connected, any model in Ollama is available, both local ones and models on Ollama's cloud, and a local model gets picked on the Settings page. Ollama says telemetry is disabled by default and prompts don't go to Anthropic or Ollama, under a Zero Data Retention policy across its models and services. Turning Claude off in Ollama restores the previous setup.