Building and Deploying .NET AI Agents with Amazon Bedrock AgentCore

This article demonstrates how to build and deploy .NET AI agents using Amazon Bedrock AgentCore, a managed service for running production AI agents with built-in scaling, session routing, and conversation memory.

• AWS.AgentCore.Hosting library connects .NET agent code to Bedrock AgentCore with HTTP communication and session management
• Two developer experiences: source generator (focus on agent logic) and extension methods (more ASP.NET Core control)
• Support for multiple model providers through IChatClient interface with Bedrock, OpenAI, Anthropic, and Ollama
• Stream responses in real-time using IAsyncEnumerable for better user experience on detailed queries
• AgentCore Memory service automatically manages conversation history per session across invocations
• Middleware support for logging, guardrails, and custom pre/post-processing without modifying core logic
• Aspire integration enables local development with embedded runtime emulator and in-memory services
• AWS.AgentCore.Testing package provides standalone emulator servers for integration testing
• Native AOT support for minimal cold start times with source-generated JSON serialization
• dotnet aws deploy tool automates container building, ECR push, and Bedrock AgentCore provisioning

Developers can now build production-ready AI agents in .NET with managed infrastructure, local testing, and one-command deployment to AWS.