Skip to main content

​
Purpose

Starcat does not resell model tokens. Configure a provider and API key in Settings → AI; requests go to that endpoint. This is the shipped path. AI features still require Pro.
Settings → AI: add provider, key, test connection

Screenshot TODO · images/screenshots/subscribe/byok.webp

​
How to set it up

Quota stays on the provider account. Chat / Embedding can be mapped per task. Example providers: OpenAI, OpenAI-compatible, DeepSeek, Gemini, OpenRouter, Ollama (local service required). The in-app list wins.
  1. Pro is active
  2. Settings → AI: add a provider, key / Base URL
  3. Map summary, chat, Embedding, and similar
  4. Test the connection
Details: AI configuration. Keys are AES-GCM encrypted in the macOS Keychain. Starcat does not upload them in cleartext to its own backend. Set a spend cap in the provider console.

​
Notes

You are not locked to OpenAI. The key stays on this Mac; traffic goes to your endpoint. See Privacy. Chat works but RAG does not: Embedding is usually empty.
Last modified on August 25, 2026