04.7-a2a-server

Chapter 4.7 -- Exposing an Agent over A2A

Companion code for Generative AI in .NET, Chapter 4 section 4.7.3 ("Exposing Your Agents via A2A").

Hosts a ChatClientAgent behind MapA2A("/a2a") so any A2A-compatible client (Python, .NET, TypeScript) can call it over HTTP using the agent card protocol.

Preview package: this sample depends on Microsoft.Agents.AI.A2A, currently published as a preview on NuGet. The pin in Directory.Packages.props is 1.3.0-preview.260423.1. Bump it as new previews ship.

Run it

ollama pull phi4-mini
dotnet run --project samples/ch04-agent-framework/04.7-a2a-server

The agent card is served at http://localhost:5182/a2a/.well-known/agent-card.json. Discover it from a remote agent or curl it directly:

curl http://localhost:5182/a2a/.well-known/agent-card.json

Pair this with A2AAgent.ConnectAsync(...) from a separate program -- see Chapter 4.7.2 for the consumer side.

Manuscript reference

• Manuscript/Chapter-04.md, section 4.7.3.
• Figure 4.9 ("A2A Multi-Agent Architecture").

Prerequisites

• .NET 9 SDK, Ollama with phi4-mini.