Continue configuration formats and setting names vary by release. In the current provider schema or settings UI, configure an OpenAI-compatible provider with these values:
| Setting | Value |
|---|---|
| Provider | openai or the release's OpenAI-compatible provider |
| API base URL | https://api.ai-router.dev/v1 |
| API key | an AI ROUTER API key supplied through a local secret or environment variable |
| Model | a current id returned by authenticated GET /v1/models |
For Continue releases using the current config.yaml schema, the equivalent
local configuration has this shape. Replace the model placeholder locally and
keep the API key in Continue's secret resolution path.
name: AI ROUTER
version: 0.0.1
schema: v1
models:
- name: AI ROUTER selected model
provider: openai
model: MODEL_ID_FROM_AI_ROUTER
apiBase: https://api.ai-router.dev/v1
apiKey: ${{ secrets.AI_ROUTER_API_KEY }}MODEL_ID_FROM_AI_ROUTER must be an actual current ID returned by /v1/models
or shown in the dashboard. Put AI_ROUTER_API_KEY=your_key in a Continue-supported
local .env file, such as ~/.continue/.env, and do not commit that file or a
plaintext key to a project-level configuration file.
- Use a dedicated test key and a single current model ID.
- Run a short, non-streaming prompt first.
- Verify that the key's usage is visible in the AI ROUTER dashboard.
- Test streaming, tool calls, and autocomplete independently if your workflow relies on them.
If automatic model discovery is incomplete, use the manual model-ID field to
enter a current ID discovered through /v1/models or the dashboard. The
configured base URL must end with /v1.
Related: Continue's official OpenAI-compatible provider documentation, client integration overview, and BYOK smoke-test checklist.