Launch the NVIDIA NeMo Agent Toolkit API Server and User Interface

Why it matters

An interaction surface is only as good as the endpoint behind it. This tutorial puts a real chat surface in front of a served workflow, so you can see which choices belong to the server and which to the client.

What to practise

Start the simple calculator workflow with nat serve, confirm it answers a curl request to /generate, and then point the NeMo Agent Toolkit UI at it. In the UI’s settings, switch the HTTP endpoint between Chat Completions and Generate, each streaming and non-streaming, and try the WebSocket schema too. Watch which modes show intermediate steps while the agent works and which return only the final answer.

How it fits

The endpoint and streaming choices put 1.1.1 into practice: streaming changes what the user sees during a run, not whether the run happens in the background. The intermediate steps the UI renders are the raw material 1.1.2 turns into concise status, tool outcomes and recoverable errors.