# Six seconds. Then under one.

> Streaming does not make it faster.

- **Format:** short video, 9 steps, ~51 seconds
- **Topic:** Streaming Microsoft Foundry agent responses from ASP.NET Core to Flutter with server-sent events: the run loop handles deltas and resumes after tool rounds, and the wait becomes legible.
- **Author:** Suthahar Jegatheesan (MSDEVBUILD)
- **Category:** Azure · Azure
- **Published:** 2026-06-15
- **Tags:** microsoftfoundry, azureai, aiagents, dotnet, flutter, azure, aicoding, msdevbuild
- **Canonical URL:** https://blog.msdevbuild.com/shorts/foundry-streaming-first-token/

---
## What you'll learn

- What streaming changes and what it does not
- How tool rounds fit a streaming loop
- How Flutter should read SSE

## Understand it one step at a time

### 1. Send, wait, send again

Four seconds of nothing, and the second tap lands.

### 2. The agent was fine

search_menu had already returned three dishes.

### 3. What streaming changes

Same total time. What changes is when the first character appears.

### 4. The streaming loop

Consume updates, run tools, resume after each round.

### 5. An SSE endpoint

data: lines, a blank line between events, one header for proxies.

### 6. Flutter, line by line

Use a line decoder. Splitting on \n yourself breaks mid-line.

### 7. Close the stream

Leaving the screen must cancel the subscription.

### 8. Under a second

Time to first token under one second; the tool pause is a status line.

### 9. Not faster, legible

Same run loop with deltas, an SSE endpoint, and a Flutter client that reads whole lines.

---

## The takeaway

**Not faster. Legible.**

Time to first token drops from about six seconds to under one. The agent is no faster.