Amazon Bedrock AgentCore: Build & Deploy any AI Agent on AWS

Explain how Amazon Bedrock Agent Core helps move agentic AI apps from proof of concept to production, using a dream vacation planner scenario.

Build a personal vacation planner app with the crew AI framework for AWS Bedrock deployment, featuring travel research and summarizer agents that search the internet and deliver a final report.

Set up the development environment with Visual Studio Code, Python 3.11.9, and AWS CLI; configure IAM with admin access in US East two to start building the agent app.

Build a vacation planner app with crew AI on AWS Bedrock, install uv, generate boilerplate, configure agents and tasks, and manage knowledge and env files.

Build a vacation planner app with two agents on Amazon Bedrock; modify boilerplate, configure yaml tasks, and run research and summarization workflows.

Implement a vacation planner CrewAI app by configuring API keys, tasks, and agents, running the project, and generating a detailed London itinerary report using Amazon Bedrock.

Build a Streamlit user interface for the crew ai vacation planner, configure Bedrock lm for agents, and deploy an interactive itinerary planner that researches destinations and outputs detailed itineraries.

Learn how to deploy an agent AI app on the agent core runtime, from local build to ecr deployment and bedrock agent core integration with endpoints.

Prepare for AgentCore deployment by installing VS Code, an AWS account, AWS CLI with an IAM role, and Docker Desktop.

Deploy a Creo AI Strands or Langraph vacation planner to the bedrock agent core runtime by adding the agent core runtime decorator, spinning up web server, and testing with curl.

Create a Dockerfile for the agent core runtime, build a cross-platform container image with docker buildx, and push the image to an AWS ECR repository via login and deploy steps.

Deploy your vacation planner agent to the Bedrock Agent Core runtime from an ECR image, configure an http endpoint, and test via the test endpoint using the Boto3 invocation code.

Expose the post endpoint of the agent core runtime via AWS Lambda and API Gateway, using boto3 to invoke the Bedrock Agent Core runtime and return the agent's JSON response.

Access the AgentCore runtime via a rest api built with api gateway and a lambda function; deploy prod and use post to invoke the vacation planner resource with cors.

Deploy your vacation planner app with a Streamlit frontend, replace the API endpoint, install Streamlit, and run the app to view the local interface powered by Amazon Bedrock Agent Core.

Learn Bedrock Agent Core observability to trace, debug, and monitor agent performance in production using CloudWatch metrics and logs, plus JNI observability with OpenTelemetry Adot.

Enable agent core observability using AWS OpenTelemetry distribution, update requirements and Dockerfile with environment variables, then deploy to ECR and enable transaction search and model invocation logging in CloudWatch.

Amazon bedrock agent core identity handles authentication and authorization between users, agents, and tools. Learn about token exchange, OAuth and IAM integration, outbound access, and observability across platforms.

Explore the Amazon Bedrock AgentCore gateway, integrated with the agent core identity to access tools securely. Discover semantic search-driven tool discovery, MCP protocol conversion, and observability enabling robust gateway operations.

Implement the Amazon Bedrock agent core gateway and identity using OAuth and Cognito, enabling a vacation planner to fetch enterprise data via a Lambda function.

Create a DynamoDB travel packages table with city as the partition key, configure a Lambda with DynamoDB access, and draft an Open API schema for the agent core gateway.

Configure a Bedrock Agent Core gateway with semantic search, enabling inbound JWT-based auth via Cognito and OpenID configuration, and define a travel agent lambda target with an inline schema.

Demonstrates hands-on gateway invocation for Amazon Bedrock AgentCore using Python, Cognito authentication, and token handling to fetch tools via json rpc 2.0.

Invoke tools through the bedrock agent core gateway with a json-rpc payload, supplying gateway url, access token, tool name, and city input for Mumbai, and IAM role permissions.

Add a lambda tool to a vacation planner agent to fetch travel packages from DynamoDB via the agent core gateway, using Cognito authentication.

Deploy changes to the agent core runtime by building and pushing a docker image to ECR, deploy via Bedrock agent core service, and test travel packages for Mumbai from DynamoDB.

Explore bedrock agent core memory, covering short-term memory, long-term memory with vector embeddings, and memory strategies like semantic summaries and user preferences for AWS agent apps.

Implement Bedrock agent core memory in your agent app by creating memory, managing memory id, session id, and actor id, and retrieving past events to append to new prompts.

Deploy memory to the Amazon Bedrock AgentCore runtime by building and pushing a vacation planner memory image to ecr, then update hosting and validate invocations via api gateway.

Learn how AI agents driven by large language models decompose complex tasks, use tools, harness memory, and apply guardrails and communication to enable sophisticated vacation planning.

Learn to design and deploy ai agents on AWS Bedrock using a cloud-agnostic, multi-agent framework. Build travel research and summarizer agents, using llm brains and tools, with sequential processes.

Explore how AI agents perform tasks, use tools, collaborate in crews, and maintain memory, guided by defined roles, goals, and backstories across sequential or hierarchical processes.

Amazon MCP standardizes how applications provide context to LLMs using a model context protocol, connecting data sources via client–server transport and tools, resources, and prompts.