Raycast
AI assistant that runs local models through Ollama, keeps conversations on your computer, and supports offline chat or an Ollama server you control.
Raycast is a proprietary desktop launcher that lets you use models running through Ollama within its AI assistant. It's for people who want private, offline conversations while keeping access to Raycast's chat and AI commands. Local model requests go directly to Ollama on your computer, without sending your conversations to Raycast or third parties. Local model access is paid.
Downloaded models work without an internet connection. Raycast detects models you've installed in Ollama and puts them in the same picker as its other AI models. You can use them in Quick AI, AI Chat, and AI Commands, so local inference fits into several parts of the app rather than a separate chat window.
What you can do depends on the model. Vision models accept image attachments, models with tool support can use AI Extensions, and thinking models can display their reasoning. Response speed and memory use depend on your hardware and model size. Smaller models respond faster; larger ones trade speed and memory for better answers.
Raycast can also connect to Ollama on another machine, such as a home server with a larger GPU. In that case, requests go to that server rather than staying on the computer running Raycast.
Some features still use cloud AI. Server-side tools such as image generation send that step to a Raycast AI model and require internet access, while the rest of a local conversation stays on your machine. The Extension AI API and Emoji Search also continue to use Raycast AI rather than local models.
