Raycast

A fast macOS launcher whose AI can run on any OpenAI-compatible endpoint with your own key.

What Raycast is

Raycast is a fast launcher and productivity hub for macOS, used to start apps, manage the clipboard and windows, and run AI commands. Its Custom Providers feature lets you add any OpenAI-compatible endpoint, so you can run Raycast AI on TokenMix with Claude, GPT or DeepSeek on your own key.

Before you start

• Install Raycast from raycast.com. It runs on macOS.
• Sign up at tokenmix.ai, open the Console, go to API Keys, and create a key that starts with sk-tm-.

Turn on Custom Providers

Custom Providers is an experimental feature, so you enable it first.

1. Open Settings, go to AI, then open Experiments.
2. Turn on Custom Providers, also called Bring Your Own Model.

Connect TokenMix

1. Still under Settings and AI, find the Custom Providers section and open its config file (providers.yaml).
2. Add a provider. The fields are id, name, base_url, api_keys and models. Use the TokenMix base URL https://api.tokenmix.ai/v1 and list at least one model:

- id: tokenmix
  name: TokenMix
  base_url: https://api.tokenmix.ai/v1
  api_keys:
    - sk-tm-your-key
  models:
    - id: claude-sonnet-4.6
      name: Claude Sonnet 4.6

3. Save the file. Field names follow whatever template your Raycast version generates in providers.yaml.

Check it works

Open Raycast AI, pick your TokenMix model, and ask a short question. A reply means it is connected. You can see the spend in the TokenMix Console.

Common questions

Where is Custom Providers: it lives under Settings, AI, Experiments, and is off by default. It is an experimental feature, so the exact wording may change between Raycast versions.

No models show up: make sure each model is listed under models in providers.yaml with a valid TokenMix model id.

Which models to pick: for quick launcher actions, fast models like Claude Haiku 4.5 and DeepSeek V4 Flash work well; use Claude Sonnet 4.6 or GPT-5.4 for heavier tasks.