OpenAI Responses

Revision 1 · Updated 2026-11-12 · 1 min read

Call the Responses API through Runtoken.

Prerequisites#

Before you start, prepare:

• A platform API key created in the console.
• The exact ID of a model supporting OpenAI Responses.
• A server-side HTTP environment; the examples use curl.

Endpoint and authentication#

Send JSON with Authorization: Bearer YOUR_PLATFORM_KEY to the following endpoint.

The OpenAI SDK base URL is https://api.runtoken.net/v1.

Send a request#

The example submits a text input and requests streamed output.

curl -N 'https://api.runtoken.net/v1/responses' \
  -H 'Authorization: Bearer YOUR_PLATFORM_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "YOUR_MODEL_ID",
    "input": "Hello!",
    "stream": true
  }'

Replace these placeholders before running the example.

curl -N disables output buffering so you can see streamed data as it arrives.

Manage conversation context#

Keep the context your application needs and send it with each request. Do not rely on a response ID to recover a previous turn.

Compact context#

The separate Compact endpoint accepts a request with an input array.

curl 'https://api.runtoken.net/v1/responses/compact' \
  -H 'Authorization: Bearer YOUR_PLATFORM_KEY' \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "YOUR_MODEL_ID",
    "input": [{ "role": "user", "content": "Hello!" }]
  }'

Replace these placeholders before running the example.

Use this request only with a model that supports Compact; Responses support alone does not establish Compact support.

Parameters and streaming#

Tools, structured output and multimodal input require model-specific confirmation before you add them to a request.

Channel switching stops once output begins. If the stream is interrupted, keep the partial result and handle any retry as a new request; billing may need review.

Next steps#

See OpenAI Chat for the message-based endpoint, or the integration guide for common routing and billing rules.