In the Open WebUI administration area, add an OpenAI-compatible connection when your installed version provides one. The exact menu labels vary by release.
Use these connection values:
- API base URL:
https://api.ai-router.dev/v1 - API key: an AI ROUTER API key
- Model: a current
idreturned by authenticatedGET /v1/models, or the same model ID shown in the AI ROUTER dashboard
Use the exact API base URL above, without a trailing slash, rather than
https://ai-router.dev. Open WebUI appends endpoint paths such as /models, so
the configured URL must already include the single /v1 suffix.
- Add a dedicated, low-scope test key through the connection's secret field.
- Refresh the model list, or add a current ID discovered through
/v1/modelsmanually if the UI supports manual entry. - Send one short, non-streaming message.
- Confirm the API key's usage in the AI ROUTER dashboard before adding more users or enabling additional features.
Never place a shared production key in a public workspace export or a committed configuration file. Test streaming, file features, and tool calls independently because OpenAI-compatible endpoints can support different feature subsets.
Related: Open WebUI's official OpenAI-compatible provider documentation, client integration overview, and BYOK smoke-test checklist.