news
Speeding up agentic workflows with WebSockets in the Responses API
April 22, 2026
OpenAI’s Codex agent loop now uses WebSockets and connection-scoped caching in the Responses API to cut API overhead and improve model latency. The change matters because agentic workflows often make many back-and-forth calls, so reducing per-call overhead can noticeably speed up interactive coding and tool-using agents.
A deep dive into the Codex agent loop, showing how WebSockets and connection-scoped caching reduced API overhead and improved model latency.
Source: openai.com