All docs
Guide
Bring your own LLM key
Orbis never proxies your key. The key goes from your machine straight to your provider. You pay your provider directly.
Supported providers
- Anthropic. Sonnet is the default planner.
- Google Gemini. 2.0 Flash is cheap and fast for vision.
- OpenAI. GPT-5 and GPT-4o family.
- DeepSeek, xAI, Mistral. Competitive at lower cost.
- OpenRouter. Useful if you want one key for many models.
- Ollama. Run a local model and never leave the machine.
Picking models
Start with the defaults shipped in the onboarding wizard. Move to a stronger planner only when a scenario is complex enough to need it. Most real test flows do well on a mid-tier model.
Cost
A typical short scenario costs around half a cent to two cents. A long multi-step suite stays under a dollar. Set the cost cap in Settings to bound any surprise.
Where the key lives
The OS keychain. Keychain Access on macOS, Credential Manager on Windows, libsecret on Linux. Orbis fetches the key only at the moment of an LLM call.