Responses
OpenAI Responses-compatible endpoint — POST /v1/responses and POST /responses.
Tera Router provides two paths for the Responses API dialect:
| Endpoint | Notes |
|---|---|
POST /v1/responses | Primary path, alongside the other /v1 endpoints |
POST /responses | Root-level alias, for Responses-native clients like Codex CLI |
Request
curl http://localhost:8080/v1/responses \
-H "Authorization: Bearer sk_tr_..." \
-H "Content-Type: application/json" \
-d '{
"model": "codex/gpt-5-codex",
"input": "Write unit tests for the parseDate function"
}'The response follows the standard Responses API format, with the X-TeraRouter-Provider and X-TeraRouter-Model headers like every other endpoint.
Streaming
stream: true is fully supported — just like Chat Completions, chunks stream as SSE.
Used by Codex CLI
Point Codex at your Tera Router instance with the standard OpenAI environment variables:
export OPENAI_BASE_URL=http://localhost:8080/responses
export OPENAI_API_KEY=sk_tr_...Or even easier: open the CLI Tools page in the dashboard and use the one-click configuration that already points at your instance.
The codex provider uses OAuth
Models under the codex provider use ChatGPT/Codex accounts (OAuth) connected through the
Providers page, not regular API keys.