From Framework to Production: Running AI Agents with Amazon Bedrock AgentCore

AI agents are easy to prototype. Running them securely, observably, and at scale? That’s where things get complicated.

In this session, we move beyond agent frameworks and focus on what it takes to operate AI agents in production using Amazon Bedrock AgentCore. While frameworks like Strands provide modular, composable agent architectures, AgentCore delivers the managed runtime, identity, memory, gateway, and observability services needed for real-world deployments.

We’ll explore:

• Secure agent runtimes with session isolation and VPC support
• Built-in identity and fine-grained authorization
• Memory services for short- and long-term context
• Gateway integration for MCP-compatible tools
• Native support for Agent-to-Agent (A2A) communication
• Observability and tracing with CloudWatch and OpenTelemetry

Through practical architecture examples, we’ll show how to take an open-source Strands agent and deploy it on AgentCore — moving from proof-of-concept to production-ready AI systems.