Skip to main content

Fetch or resume an agent response

GET 
/agent/:id/responses/:response_id

Re-opens a previously-started response. Behaviour depends on starting_after:

  • Resume / replay (SSE): with starting_after=<sequence_number> (and Accept: text/event-stream), the platform replays every event whose sequence_number > starting_after and, if the response is still in progress, continues live as new events arrive. Re-resuming yields the same sequence_number for the same logical event, so the cursor is stable. starting_after past the last event closes the stream immediately with no events (a lightweight "no new events" probe).
  • Terminal metadata read (JSON): when starting_after is absent and the response has already terminated, the platform returns the cached terminal OpenAI Response object as JSON.

See the Responses resume guide for the resume protocol and the full SSE taxonomy.

Request​

Responses​

When resuming/replaying: an SSE stream of the remaining events. When reading a terminated response without starting_after: the terminal OpenAI Response object as JSON.