# Build Production-Ready AI Agents in Minutes with Amazon Bedrock AgentCore and Kiro Source: https://www.youtube.com/watch?v=dBhdowXEniw Recap page: https://rapidrecap.app/video/dBhdowXEniw Generated: 2025-12-15T21:03:40.043+00:00 --- ## Quick Overview The video successfully demonstrates building and deploying a production-ready AI agent using Amazon Bedrock AgentCore and the Kiro IDE, showing how to initialize a project, configure an AWS-backed agent with memory capabilities (both short-term and long-term), test it locally via a REST API endpoint, and finally deploy it using the AgentCore CLI, all within minutes. **Key Points:** - The demonstration uses the Kiro IDE to quickly set up a new project using the 'agentcore-get-started' template (0:30). - The process involves configuring the AgentCore MCP Server within the 'mcp.json' file, which enables communication between the agent and the necessary infrastructure (0:51). - A simple 'hello_agent.py' is created and tested locally using 'uv run hello_agent.py', confirming the agent's basic functionality (2:29). - Memory capabilities are added by importing 'BedrockAgentCoreMemory' and using 'MemoryClient' to create both Short-Term Memory (STM) and Long-Term Memory (LTM) resources (10:49). - The agent is configured for deployment using 'uv run agentcore configure --entrypoint hello_agent.py' (6:17), which sets up AWS resources like an Execution Role and ECR Repository (7:25). - Deployment is finalized using 'uv run agentcore launch', which builds ARM64 containers in the cloud with CodeBuild and confirms deployment success with an ECR URI and next steps like 'agentcore invoke' (8:36). - The video concludes by successfully invoking the deployed agent via a curl command, confirming the memory functionality is retained in the cloud deployment (9:08). ![Screenshot at 0:36: The Kiro IDE interface is displayed, showing the main dashboard with options to 'Let's build' using 'Vibe' or 'Spec' modes, alongside a list of available commands for starting a new session.](https://ss.rapidrecap.app/screens/dBhdowXEniw/00-00-36.png) **Context:** This tutorial, presented by Erik Hanchett, Senior Developer Advocate, focuses on accelerating the development and deployment cycle for AI agents built on Amazon Bedrock AgentCore. The primary tool utilized is the Kiro IDE, which integrates CLI commands and configuration management to streamline setting up complex agent infrastructure, including memory features, making the process accessible even for production readiness. ## Detailed Analysis The video walks through building a production-ready AI agent using Amazon Bedrock AgentCore, leveraging the Kiro IDE for rapid setup. The presenter first initializes a project using the 'agentcore-get-started' template (0:30). Key initial steps involve configuring the necessary MCP Servers in 'mcp.json' (0:51). A simple 'hello_agent.py' file is created and tested locally using 'uv run hello_agent.py' and curl commands to verify basic invocation (1:29, 4:46). The core feature demonstrated is the addition of memory capabilities, where 'BedrockAgentCoreMemory' is imported, and a setup script ('setup_memory.py') is run to create both Short-Term Memory (STM) and Long-Term Memory (LTM) resources, including semantic strategies for fact extraction (10:49, 11:14). The presenter then uses the AgentCore CLI to configure the agent for deployment ('uv run agentcore configure --entrypoint hello_agent.py') (6:17), which automatically handles resource creation like IAM Execution Roles and ECR repositories (7:25). Finally, the agent is deployed to AWS using 'uv run agentcore launch' (8:27). The output confirms deployment success, providing an ECR URI and instructions to invoke the deployed agent via the AWS endpoint (8:36). The session closes by demonstrating the successful invocation of the cloud-deployed agent using a curl command, proving memory persistence across deployment (9:08). ### Project Initialization and MCP Setup - Getting started with 'agentcore-get-started' - Configuring 'mcp.json' to connect to the Bedrock AgentCore Runtime MCP Server - Verifying initial files created (0:30, 0:51) ### Local Testing of Base Agent - Creating 'hello_agent.py' with a simple invocation function - Testing locally via 'uv run hello_agent.py' and verifying response via curl command (1:29, 4:29) ### Adding Memory Capabilities - Importing 'BedrockAgentCore.memory' - Running 'setup_memory.py' to create STM and LTM resources using specific strategies (e.g., 'userPreferenceMemoryStrategy') - Exporting MEMORY_ID as an environment variable for persistence (10:49, 11:32) ### Agent Configuration and Deployment - Running 'uv run agentcore configure --entrypoint hello_agent.py' to set up AWS resources (Execution Role, ECR) - Confirming IAM authorization is used by default (6:17, 7:06) ### Deployment and Final Invocation - Executing 'uv run agentcore launch' which uses CodeBuild for cloud deployment - Receiving deployment success confirmation with ECR URI - Testing the deployed agent via 'agentcore invoke' command (8:27, 8:46) ![Screenshot at 0:01: Erik Hanchett, Senior Developer Advocate, introduces the topic of getting started with scalable Bedrock AgentCore agents.](https://ss.rapidrecap.app/screens/dBhdowXEniw/00-00-01.png) ![Screenshot at 0:36: The Kiro IDE main screen displaying the 'Let's build' interface and available agent core commands like 'Open chat' and 'Show All Commands'.](https://ss.rapidrecap.app/screens/dBhdowXEniw/00-00-36.png) ![Screenshot at 0:51: The 'mcp.json' configuration file open, showing the defined 'mcpServers' entry pointing to the 'aws-labs.amazon-bedrock-agentcore-mcp-server' \(51\).](https://ss.rapidrecap.app/screens/dBhdowXEniw/00-00-51.png) ![Screenshot at 2:09: The Kiro chat panel showing the agent's successful response to the initial query, including logs indicating the agent is ready.](https://ss.rapidrecap.app/screens/dBhdowXEniw/00-02-09.png) ![Screenshot at 4:45: The terminal showing a curl command being constructed to invoke the locally running agent at 'http://localhost:8080/invocations' \(4:45\).](https://ss.rapidrecap.app/screens/dBhdowXEniw/00-04-45.png)