Persistent AI agents. Up to 40% faster.
Every agent turn, you're resending the full context. Again. That overhead compounds fast. WebSocket Mode for the Responses API keeps a persistent connection, sends only incremental inputs, and cuts end-to-end latency by up to 40% on heavy tool-call workflows.
WebSocket mode uses previous_response_id chaining to continue sessions, allowing developers to send only new input items while the API maintains full context. This enables stateful multi-turn interactions without resending prior conversation history.
WebSocket mode maintains a persistent connection to the Responses API and sends only incremental inputs (new tool outputs and messages) on each turn, eliminating per-turn connection overhead. For workflows with 20+ tool calls, this delivers approximately 40% faster end-to-end execution compared to standard HTTP requests.
WebSocket mode is compatible with both Zero Data Retention (ZDR) and store=false options, allowing developers to reduce latency without incurring storage costs. The incremental input pattern also reduces token overhead per turn.
WebSocket mode is optimized for long-running, tool-call-heavy workflows such as agentic coding or orchestration loops. The persistent connection and incremental input pattern reduce continuation overhead, making it ideal for multi-step agent reasoning and tool execution chains.