Skip to main content

Ollama

Ollama runs open-source models locally on your Mac (or another machine you control). Personas that use the ollama provider talk to Ollama’s OpenAI-compatible HTTP API with local model names, for example llama3.2 or mistral.

No cloud API key is required for the default local install.

Install and run Ollama​

  1. Download and install Ollama from ollama.com.
  2. Start Ollama (the menu bar app or ollama serve).
  3. Pull a model, for example:
ollama pull llama3.2
  1. Confirm the local API is up (default: http://localhost:11434).

Configure in Toby​

Open Toby.app → Settings → AI → Ollama:

  1. Set Base URL to your Ollama OpenAI-compatible endpoint (Toby defaults to http://localhost:11434/v1).
  2. Leave API Key empty unless your Ollama instance requires one.

For a persona, set AI Provider to ollama and choose a model you have pulled (or type a model name Ollama knows).

Tips​

  • Model ids match ollama list names (for example llama3.2, not a cloud slug).
  • If Ollama runs on another host, point Base URL at that machine’s reachable address and ensure the port is open.
  • For more context on choosing providers, see AI providers overview.