SSE

Streaming output pushes tokens chunk by chunk via Server-Sent Events (SSE) or HTTP chunked encoding, and the frontend renders each increment immediately upon receiving it.

Streaming responses mean the model returns content incrementally via SSE or a data stream as it generates, allowing the frontend to render in real time.

SSE (Server-Sent Events) is a server push technology defined in the HTML5 specification that only supports a one-way stream from server to client; MCP in SSE mode adopts a dual-channel design, where the client sends requests via HTTP POST and the server's responses and notifications are sent back through the SSE stream.

Streaming output relies on HTTP chunked transfer or SSE (Server-Sent Events); if the frontend uses ordinary fetch to read response.json(), it waits for the entire stream to finish before parsing, causing the interface to stay blank for a long time; the correct approach is to read the ReadableStream of response.body and decode it chunk by chunk.

Session communication is done through an event stream: the frontend sends user.message events, and the backend listens to the SSE stream and pushes the agent's responses back to the browser in real time.

When connecting directly via the Chat API, access credentials should be kept on the server side, long-lived Tokens should not be exposed in the browser, and attention should be paid to how streaming responses (SSE) are integrated.

In the MCP specification, Sampling, Roots, and Logging have been marked as deprecated and can still be used for at least 12 months; the long-running Tasks capability has been moved out of the core to become an official extension; SSE is the deprecated legacy transport method—local processes use Standard IO, and remote services use Streamable HTTP.

In long tasks, simple polling is inefficient, and SSE or WebSocket are more common choices for streaming progress updates, but backpressure and reconnection must be handled to avoid duplicate or out-of-order events.

A2A's task model is designed for long-running scenarios: a task delegation produces a Task object with a unique ID whose state transitions among submitted, working, input-required, completed, failed and others, during which progress updates can be streamed via SSE, and push notification callbacks are also supported.

The AI SDK's streaming responses typically rely on SSE or similar streaming HTTP protocols at the underlying level.