Add or replace a provider key
- Open Settings → Models and expand API Keys.
- Find the intended provider row, such as OpenAI, Anthropic, Meta, Google, xAI, Cursor, or OpenRouter.
- Enter the key in its secret field and save settings. Review any validation error; Anthropic key saving includes verification.
- Reveal a stored key only when you need to inspect it, and clear it only when you intend to remove that connection.
- Return to the feature or harness that uses the connection and select the matching route.
Keys are managed on the device through Rudr’s secret-handling path. Do not paste them into prompts or diagnostic descriptions. A configured indicator is not proof that every model or endpoint is authorized.
Configure Azure OpenAI or Foundry
Expand Azure OpenAI on the Models page. Review the key, Base URL, API version, Endpoint mode, and configured model/deployment mappings. Choose whether the connection should route the applicable OpenAI or Claude models, then save.
Use the actual endpoint and deployment information from your account. A friendly display name is not necessarily the deployment identifier. After saving, select the explicit Foundry route in a supporting Code harness and check its model discovery.
Manage the model catalog
Use Search models and Refresh model catalog. Check the last successful update; Using the last saved catalog means the current refresh failed. Visibility toggles control what is offered, rather than granting entitlement to a provider model.
Under Custom models, choose Add model when you have the correct provider and identifier. Use the row’s Edit or Delete action to maintain an existing custom entry. Removing a custom entry affects selection in Rudr; it does not delete a deployment at the provider.
Grok Build’s XAI_API_KEY environment route is separate from the xAI field used by voice and transcription. See Grok setup and authentication recovery for route-specific problems.