Skip to content

AgentCore CLI

The easiest way to deploy and manage AI agents on Amazon Bedrock AgentCore Runtime

A sophisticated command-line interface that brings environment-first architecture and single-command deployment to AI agent management. Deploy from container to runtime in one command, manage multiple environments, and scale from development to production with confidence.

Note

Amazon Bedrock AgentCore is currently in preview release and is subject to change.

🚀 Core Values

  • | Environment-First Design

    Complete isolation between dev, staging, and prod environments. Each environment maintains its own agent runtimes, endpoints, and configurations with independent AWS regions.

  • | Agent Lifecycle Management

    Immutable versioning system where every update creates a new version. Endpoints can point to any version, enabling safe rollbacks and blue-green deployments.

  • | Container-Native

    Built for Docker workflows with Amazon Bedrock AgentCore Runtime. Works with any agent framework including LangGraph, CrewAI, Strands Agents, and custom implementations.

  • | Infrastructure as Code

    CloudFormation templates for ECR repositories, IAM roles, and Cognito resources. Consistent, repeatable deployments with full resource lifecycle management.

  • | Configuration Sync

    Synchronize configuration between local development and cloud (AWS Parameter Store) with intelligent drift detection using DeepDiff.

  • | Security-First

    Integrated IAM role management and Cognito authentication. Every agent gets appropriate permissions and secure access patterns with built-in identity management.

⚡ Quick Start

Bash
1
2
3
4
5
# Interactive setup wizard
agentcore-cli init

# Or automated setup
agentcore-cli init --no-interactive --region us-west-2 --environment dev
Bash
1
2
3
4
5
6
7
8
9
# Creates ECR repo, builds image, creates IAM role, deploys runtime
agentcore-cli agent create my-chat-bot

# With custom settings
agentcore-cli agent create ml-processor \
  --environment prod \
  --image-tag v1.0.0 \
  --dockerfile ./production.Dockerfile \
  --build-args API_KEY=secret123
Bash
1
agentcore-cli agent invoke my-chat-bot --prompt "Hello, how are you?"
Bash
1
2
3
4
5
6
7
8
# Create production environment
agentcore-cli env create prod --region us-east-1

# Switch to production
agentcore-cli env use prod

# Deploy to production
agentcore-cli agent create my-chat-bot --environment prod

📚 Documentation Sections

  • | Getting Started

    Install the CLI, set up your AWS credentials, and deploy your first agent in minutes.

  • | User Guide

    Learn about environments, agent lifecycle, containers, configuration, and resource management.

  • | Architecture

    Understand environment-first design, security model, and AgentCore Runtime capabilities.

  • | CLI Reference

    Comprehensive reference for all commands, options, and arguments.

  • | Tutorials

    Step-by-step guides for common workflows and advanced features.

  • | Examples

    Ready-to-deploy examples for different agent frameworks.

  • | Troubleshooting

    Common issues, error messages, and debugging tips.

  • | API Reference

    Auto-generated documentation for the Python library.

🧩 Architecture Overview

The AgentCore CLI uses an environment-first architecture that provides clean separation between development, staging, and production environments:

graph TD
    A[Environment] --> B[Agent Runtime]
    B --> C[Versions V1, V2, V3...]
    B --> D[Endpoints DEFAULT, production, staging]
    D --> C
    C --> E[Container Images]
    E --> F[ECR Repository]
    B --> G[IAM Role]
    A --> H[Cognito Auth]

🛠️ Development Setup

Prerequisites

  • Python 3.11+
  • Docker installed and running
  • AWS CLI configured with appropriate permissions
  • Access to Amazon Bedrock AgentCore Runtime (Preview)

Installation

Bash
1
2
3
4
5
6
7
# Install from PyPI (when published)
pip install agentcore-cli

# Or install from source
git clone https://github.com/yourusername/agentcore-cli.git
cd agentcore-cli
uv sync  # Using uv for dependency management

Built with ❤️ for the AI developer community

Deploy smarter, not harder. One command, infinite possibilities.