Speeding up agentic workflows with WebSockets in the Responses API

In three linesOpenAI optimizes agentic workflows using WebSockets in the Responses API: reduced model latency and API overhead through connection-scoped caching. Improvements documented on the Codex agent loop.

Summary generated by Claude — human-verified