Google Cloud has introduced streaming support for API Gateway in public preview. The change extends the service beyond traditional request-response traffic and targets applications that need progressive model output, agent status updates, tool-calling connections, or other real-time interactions.
Supported patterns include Server-Sent Events, chunked transfer encoding, HTTP response streaming, WebSockets, and gRPC. Connections may remain open for up to one hour, after which a client must reconnect to continue communicating. Developers do not need to alter existing API specifications or add special annotations.
Streaming endpoints retain API Gateway controls such as API key enforcement, JWT validation, service account authentication to backends, request routing, path templating, Cloud Logging, and Cloud Monitoring. The service also offers model routing through a single OpenAI-compatible endpoint for Gemini, Claude, or OpenAI gpt-oss models. Google says combining routing with streaming lets clients use a familiar interface while administrators keep centralized access policies.
The preview also introduces predictable gateway hostnames based on the gateway ID, project number, and region. There are limitations to consider: existing gateways cannot be converted in place, and request-latency metrics measure the full stream duration rather than time to first byte. Developers must enable the feature with the --enable-streaming flag when creating a gateway.