Skip to content

Open models

Claude Desktop picks up an Ollama gateway toggle

Written by an AI model from the linked sources and checked against them by a second model. How we work

Image: ollama.com

Ollama can now act as a third-party gateway for Claude Desktop, so the app can run open models on your own machine or on Ollama’s cloud. The setup happens on Ollama’s side. In an August 26 post, Ollama lists three steps: open Ollama, select Claude, turn Claude on. Ollama then configures the gateway itself.

Once connected, you can pick any model in Ollama, local or cloud. Ollama suggests its cloud models for larger coding and research tasks, and local ones for running directly on your computer. To use a local model, set it on Ollama’s Settings page. Anthropic models are still available: the toggle inside Ollama switches between them and Ollama’s models.

Until now, the Ollama route was a manual job for Claude Code. Ollama 0.14.0, announced on January 16, added support for the Anthropic Messages API, the request format Claude clients use, and Claude Code could then be pointed at a local Ollama server through two environment variables: ANTHROPIC_BASE_URL=http://localhost:11434 and ANTHROPIC_AUTH_TOKEN=ollama. For the Desktop app, a switch in Ollama does the setup.

On data handling, Ollama answers “No.” to whether prompts or data are sent to Anthropic or Ollama. Telemetry is disabled by default, and Ollama says it applies a Zero Data Retention policy “across all of our models and services whether you access them through our cloud or on your local machine.”

Undoing it takes the same switch. Turn Claude off in Ollama, and Claude Desktop goes back to the setup it had before.

No comments yet. Be the first.

Join the conversation

Sign in to leave a comment. It appears after moderation.