Capabilities and compatibility
The-Code API provides a common integration surface for models from multiple providers. Availability and supported parameters depend on the models enabled for your account.
| Endpoint | Capability | Compatibility notes |
|---|---|---|
GET /v1/models | Model discovery | Authoritative catalog for the calling key, including request contracts |
POST /v1/chat/completions | Chat and tool calling | OpenAI-compatible requests; streaming where the model and requested features support it |
POST /v1/responses | Responses | Documented non-streaming subset; do not assume stateful or background support |
POST /v1/embeddings | Embeddings | Requires a model advertising embeddings |
GET /v1/requests/{public_id} | Request status | Poll asynchronous gateway requests using the same API key |
Managed video features are also available through the customer portal and documented API routes, subject to account access. They are separate from the OpenAI-compatible chat and embeddings endpoints.
Inspect the model contract
Function calling, structured output, reasoning settings, output limits, and parallel tool behavior vary by model. Read the selected model's live request contract before sending optional parameters. The same SDK method does not imply identical features across providers.
Use exact model IDs for exact routing. An alias allows administrator-managed routing. Do not silently substitute a different model when a requested model is unavailable.
See the detailed API guide, OpenAPI reference, and agent guide.