Agent Lifecycle Management¶

Work in Progress

This page is currently under development.

Overview¶

Agent lifecycle management is a fundamental aspect of AgentCore CLI that covers the creation, versioning, deployment, invocation, and deletion of agent runtimes.

Agent Concepts¶

Agent Runtime: A containerized AI application deployed on Amazon Bedrock AgentCore Runtime
Versions: Immutable snapshots of an agent at a specific point in time
Endpoints: Named pointers to specific versions (e.g., DEFAULT, production, staging)
Invocation: Sending requests to an agent through an endpoint
Lifecycle: The complete process from creation to deletion

Agent Creation¶

Create a new agent using the agent create command:

Bash
agentcore-cli agent create my-bot --dockerfile ./Dockerfile

Agent Versioning¶

AgentCore CLI uses an immutable versioning system where every update creates a new version:

Bash
# Update an agent, creating a new version
agentcore-cli agent update my-bot --image-tag v2.0.0

Endpoint Management¶

Endpoints can point to any version, enabling safe rollbacks and blue-green deployments:

Bash
# Update a specific endpoint to point to a version
agentcore-cli agent update my-bot --endpoint production --image-tag stable-v1

Agent Invocation¶

Invoke an agent with a prompt or payload:

Bash
# Invoke with a text prompt
agentcore-cli agent invoke my-bot --prompt "Hello, agent!"

# Invoke with a custom payload
agentcore-cli agent invoke my-bot --payload '{"query": "Hello", "max_tokens": 100}'

Agent Status and Listing¶

View the status of agents and list deployed agents:

Bash
# Show status of a specific agent
agentcore-cli agent status my-bot

# List all agents
agentcore-cli agent list

Agent Deletion¶

Delete agents that are no longer needed:

Bash
agentcore-cli agent delete my-bot --force

Best Practices¶

Guidelines for agent lifecycle management will be added in a future update.

Detailed Documentation¶

Complete documentation will be added in a future update.