A travel planning agent demonstrating the journey from local PoC to production-ready deployment using Amazon Bedrock AgentCore and Strands Agents SDK.

This demo showcases two implementations:

1. Local PoC Agent (travel_agent_local.py): Simple agent with inline tools for rapid prototyping
2. Production Agent (travel_agent_production.py): Enterprise-ready agent with:
   • Memory: Short and long-term context via AgentCore Memory
   • Gateway: Centralized MCP-compatible tools via Lambda
   • Identity: OAuth authentication with Cognito
   • Observability: Built-in monitoring and logging

• AWS account with credentials configured
• Python 3.10+
• Docker or Finch (for local testing)
• Model Access: Anthropic Claude 3.7 enabled in Amazon Bedrock
• AWS Permissions: BedrockAgentCoreFullAccess and AmazonBedrockFullAccess

pip install -r requirements.txt

No setup required - just run:

python travel_agent_local.py

Interact via CLI:

You: I want to fly from New York to Paris on 2025-06-15
Agent: [response]

Setup AWS resources:

python scripts/setup_memory.py
python scripts/setup_cognito.py
python scripts/setup_gateway.py

This creates:

• AgentCore Memory with USER_PREFERENCE and SEMANTIC strategies
• Cognito User Pool and Client for OAuth
• Lambda function with travel tools
• AgentCore Gateway with MCP endpoint and JWT auth
• IAM roles and SSM parameters

Test Production Agent Locally:

python travel_agent_production.py

Then in another terminal:

curl -X POST http://localhost:8080/invocations \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Tell me about Tokyo"}'

Deploy production agent to AWS:

python scripts/deploy_runtime.py

This will:

• Create IAM execution role
• Build and deploy Docker container
• Configure Memory, Gateway, and Cognito integration
• Wait for deployment to be READY

Test deployed agent:

python scripts/test_runtime.py

To delete all created resources:

# If deployed to Runtime
agentcore destroy

# Cleanup AWS resources
python scripts/cleanup_gateway.py
python scripts/cleanup_cognito.py
python scripts/cleanup_memory.py

• "I want to fly from New York to Paris on 2025-06-15"
• "Tell me about Tokyo"
• "What are the top attractions in Sydney?"
• "Find me flights from Los Angeles to Sydney next month"
• "I'm planning a trip to London, what should I know?"

• New York ↔ Paris
• London ↔ Tokyo
• Los Angeles ↔ Sydney

Additional destinations available for information: Paris, Tokyo, Sydney, London, New York

• Short-Term Memory: Conversation context within sessions
• Long-Term Memory:
  • USER_PREFERENCE: Traveler preferences and behavior
  • SEMANTIC: Factual information from conversations