Raycast

Raycast's AI assistant lets you chat with models running through Ollama on your computer. It's for people who want local AI conversations within Raycast's existing chat and command tools.

• Chat and commands: Local models work in Quick AI, AI Chat and AI Commands, and appear alongside Raycast's other models in the model picker.
• Model capabilities: Compatible models can accept image attachments, use AI Extensions or display reasoning, depending on the capabilities Ollama reports.
• Your own server: Raycast can connect to Ollama on another machine, such as a home server with a larger GPU, instead of running inference on your computer.

Once you've downloaded a model, local conversations work offline. Requests go directly to Ollama on your machine, and Raycast doesn't send those conversations to its servers or third parties. If you choose a remote Ollama server, requests go to that server. Response speed and memory needs depend on your hardware and model size.

Some capabilities still require cloud processing. Server-side tools such as image generation use a Raycast AI model for that step and need internet access, even within a local conversation. The Extension AI API and Emoji Search also use Raycast AI rather than local models.