Amazon Bedrock AgentCore Samples
CLI-led samples for developing and deploying multi-framework AI agents on AWS AgentCore.
Per-dimension scores and reasoning
Evidence shows: repository includes security scan workflows (ASH), but no explicit least-privilege principle; examples may involve AWS credential configuration, but user confirmation mechanism is not emphasized; data flow transparency is limited, only mentioned in docs; sensitive data handling not detailed; dependency security partially addressed via scan workflow, but no dependency locking or vulnerability mitigation details; external effects (e.g., deployment to AWS) documented, but rollback mechanism not covered; source attribution via contributor list and license, but publisher unverified. Deductions: lack of user confirmation, rollback, and insufficient evidence for sensitive data handling and dependency security.
Evidence shows: repository structure clear, README consistent with directory structure, but no self-consistency tests; dependency availability not explicitly stated, only listed; failure messages not documented. Deductions: lack of explicit dependency availability and failure handling.
Evidence shows: targets multiple frameworks and models, provides various scenario examples, but trigger precision (e.g., CLI commands) not detailed; environment fit (e.g., AWS configuration) documented, but not all environments covered. Deductions: insufficient evidence for trigger precision and some environment fit.
Evidence shows: clear information architecture, detailed install notes, stable naming (consistent directory structure), rich examples, explicit license (Apache-2.0), maintenance responsibility via contribution guidelines, but known limitations and version changelog not explicit. Deductions: missing known limitations and version changelog.
Evidence shows: output usability via examples and docs, high marginal value (multiple integrations), but cost-benefit not quantified. Deductions: lack of specific cost-benefit data.
Evidence shows: some claims supported by docs, but lack cross-source verification, facts and inferences not clearly separated. Deductions: insufficient verification evidence.
- Publisher identity unverified, treat source with caution.
- Dependencies not pinned, supply chain risk.
- Examples may involve AWS credentials, ensure secure configuration.
- Lack of rollback mechanism, difficult to recover after deployment issues.
What does this agent do, and when should you use it?
This is a collection of Amazon Bedrock AgentCore samples and tutorials, not a single chat-ready agent. Its material is organized into getting-started, features, end-to-end, integrations, infrastructure-as-code, blueprints, and legacy areas. It covers AgentCore Runtime, Gateway, Identity, Memory, built-in tools, observability, evaluation, and Policy, with Python and TypeScript samples. The recommended delivery path uses the AgentCore CLI for local development and deployment to Amazon Bedrock AgentCore Runtime, while end-to-end applications and blueprints combine multiple capabilities. It fits teams that have chosen AWS for agent operations and want executable implementation and deployment references.
After installing the CLI with npm install -g @aws/agentcore, agentcore create interactively scaffolds a project with a selected framework—such as Strands Agents, LangGraph, Google ADK, or OpenAI—and either Python or TypeScript. agentcore dev starts a local development server, watches file changes, and exposes a local invocation endpoint. agentcore deploy deploys the project to Amazon Bedrock AgentCore, and agentcore invoke tests the deployed agent. The CLI also exposes agentcore add memory, agentcore add identity, agentcore add evaluator, and agentcore add online-eval; Gateway samples describe converting APIs, Lambda functions, and services into MCP-compatible tools.
- An AWS engineering team needs to scaffold a Python agent, validate it locally, and deploy it to AgentCore Runtime.
- A team using Strands Agents, LangGraph, CrewAI, or LlamaIndex needs implementation references for AgentCore infrastructure.
- An agent-platform engineer needs to expose existing APIs, Lambda functions, or services as MCP-compatible tools.
- A developer needs managed Memory for a personalized agent or Identity integration for third-party applications.
- A product team needs an end-to-end reference that combines runtime, identity, tools, and observability capabilities.
What are this agent's strengths and limitations?
- Provides a defined AgentCore CLI workflow from project creation through local development, deployment, and invocation testing.
- Includes Python and TypeScript samples and explicitly targets multiple agent frameworks and model choices.
- Separates Runtime, Gateway, Identity, Memory, observability, evaluation, and Policy into focused capability examples.
- Supplies end-to-end applications, infrastructure-as-code templates, and customizable full-stack blueprints.
- Its core deployment boundary is Amazon Bedrock AgentCore Runtime, so adopting AWS is a prerequisite.
- The quick start requires an AWS account, configured credentials, two AWS managed permissions, and Bedrock model access.
- The CLI requires Node.js 20.x or later, and the Python path also requires
uv. - Some samples still depend on the previous Starter Toolkit and are undergoing migration to the AgentCore CLI.
How do you install or deploy this agent?
Prerequisites: configure an AWS account and credentials with aws configure; install Node.js 20.x or later; use uv for Python agents or Node.js for TypeScript agents; enable Anthropic Claude 4.0 model access in the Amazon Bedrock console. Required AWS permissions are BedrockAgentCoreFullAccess and AmazonBedrockFullAccess. Run:
npm install -g @aws/agentcore
agentcore create
cd my-agentHow do you use this agent?
From the created project directory, run agentcore dev to start the local development server and test its local invocation endpoint. Run agentcore deploy to deploy, then agentcore invoke to test the deployed agent. To extend the project, run commands such as agentcore add memory, agentcore add identity, agentcore add evaluator, or agentcore add online-eval, then run agentcore deploy again to sync changes.
How does this agent compare with similar options?
The repository positions the AgentCore CLI as the recommended path for creating, developing, and deploying agents, replacing the earlier Bedrock AgentCore Starter Toolkit workflow. Samples still using the Starter Toolkit are placed in legacy/, with a migration mapping in MIGRATION.md.
FAQ
Is this a single agent that I can deploy directly?
Which AWS permissions are required to deploy?
BedrockAgentCoreFullAccess and AmazonBedrockFullAccess, along with configured AWS credentials.Does it limit me to one agent framework or model?
What does it cost to use?
What happens to Starter Toolkit samples?
legacy/, and MIGRATION.md provides the mapping.