Connect custom models that use the Responses API
Custom inference endpoints now support OpenAI-compatible Responses requests alongside Chat Completions, with an API-format selector in the dashboard.
Connect a custom inference endpoint that expects the OpenAI Responses format. Choose Responses or Chat Completions when configuring the connection in the dashboard, API, or MCP. Existing connections keep using Chat Completions by default.
Responses requests use store: false, and supported endpoints can receive tool calls and declared reasoning settings through the same connection. The dashboard previews the endpoint path for the format you choose.
This is Responses-format support for custom endpoints, not support for every provider-specific extension. Anthropic Messages is not part of this connector release.