Models and providers
Connect providers, choose main and auxiliary models, and understand usage.
Choose a provider
Use /provider for connections and /model for model selection. The installed provider configuration and model picker are the current source of truth for available models.
| Provider | Command |
|---|---|
| DeepSeek | /provider deepseek |
| Kimi | /provider kimi |
| MiniMax | /provider minimax |
| Zhipu | /provider zhipu |
| MiMo | /provider mimo |
| LongCat | /provider longcat |
| OpenRouter Free | /provider openrouter-free |
| Local compatible API | /provider local |
| Custom compatible API | /provider custom |
ChatGPT Codex sign-in
The Codex connection uses device-code OAuth. Run the command, then follow the verification URL and device code displayed by Crux in your browser. This is separate from configuring an OpenAI API key.
/provider codex loginEnvironment variables
Set credentials before starting Crux if you prefer environment configuration. A provider-specific key overrides the global default. Replace the placeholder with your key and do not commit it.
- CRUX_API_KEY: global default key.
- CRUX_API_KEY_<PROVIDER>: uppercase provider name, for example CRUX_API_KEY_MINIMAX.
export CRUX_API_KEY_DEEPSEEK='your-api-key'
crux /path/to/projectAuxiliary models and reasoning
Auxiliary models handle tasks such as TLDR and session titles. Enter /auxiliary and use completion to choose an available model; /auxiliary none disables it. Use /think to adjust reasoning. Supported levels depend on the model.
/auxiliary
/think adaptive
/modelWhat usage indicators mean
Coding-plan quotas and balance displays depend on provider support. Missing data does not mean zero remaining usage. Costs are estimates based on configured prices and returned usage; provider billing remains authoritative.