Configure
Providers and models
Configure Claude Code, Codex, ChatGPT, Ollama Cloud, and API-based inference providers.
Motriz is provider-neutral at the project and flow level. Configure only the providers you want; unavailable providers stay visible so the UI can explain what is missing.
Supported providers
| Provider | Connection | Notes |
|---|---|---|
| Claude Code | Local claude CLI and existing subscription login | Auto-detected from PATH or configured with a custom CLI path. |
| Codex | Local codex CLI and existing Codex login | Runs through Codex app-server with an app-managed runtime home. |
| ChatGPT | Browser sign-in | Supports sign-in, cancel, account status, and sign-out from Settings. |
| Anthropic | API key | Optional custom base URL. |
| OpenAI | API key | Direct API access. |
| Azure OpenAI | Endpoint URL and API key | Uses your Azure resource endpoint. |
| Grok | xAI API key | OpenAI-compatible runtime. |
| Ollama Cloud | Ollama Cloud API key | Motriz supplies repository write tools and local bash in Build. |
| OpenRouter | API key | Models come from the provider catalog and typed IDs. |
| Hugging Face | Access token | Uses compatible hosted inference models. |
| Cerebras | API key | Uses compatible hosted inference models. |
The current app uses your provider subscription login or API credentials. It does not implement a Motriz Cloud credit balance, hosted account, or paid subscription gate.
Configure a CLI provider
- Install and authenticate the CLI.
For Claude, run
claude auth login. For Codex, runcodex login. - Open Settings → Providers.
Expand the provider card and check its status detail.
- Use auto-detection first.
If Motriz finds the binary, no custom path is required. Otherwise enter the absolute CLI path and Save.
- Verify model availability.
Ready models appear in the card and in model pickers.
Configure an API provider
Enter the API key in the provider card and Save. The input is cleared after saving because secrets are never read back into the UI. To rotate a key, enter the replacement. To remove it, choose Remove.
Provider settings cannot be changed while any agent run is active. Stop runs first so a single execution cannot change credentials or endpoints midway through a request.
Choose a model
The model picker groups entries by provider, prioritizes available defaults, supports search, and can show hidden models when requested. You can also type a model ID when the provider supports models not present in the seeded catalog.
Only configured providers are usable. If a model shows as unavailable, connect that provider or choose another model.
Choose reasoning effort
| Provider family | Effort choices |
|---|---|
| Claude Code and Anthropic | Low, medium, high, xhigh, max. |
| Codex and ChatGPT | Low, medium, high, xhigh; supported GPT-5.6 Sol configurations also expose Ultra. |
| Ollama Cloud | No thinking or thinking. |
| OpenAI-compatible API providers | Low, medium, high where the model supports reasoning effort. |
When you switch models, Motriz keeps the selected effort only if the new model supports it.
Understand Fast
Fast is a service tier, separate from reasoning effort. For Codex and ChatGPT it maps to the provider's priority service tier; for supported Anthropic models it maps to fast speed. Availability is model-specific.
Choose a role
Build roles include Coder, Planner, Heavy Planner, Reviewer, and System Architect. Chat, Product Design, and venture threads pin their own mode-specific personas. Custom flow steps can select a role and add instructions specific to that pipeline.
Understand usage and cost
Run timelines display token usage and estimated cost when the provider reports enough information. The dashboard's seven-day spend aggregates those run records. It is a local estimate, not a provider invoice, and some providers may return no cost data.