Conclusion and next steps
You've built an AI agent that combines real-time behavioral data from Snowplow Signals with persistent memory from AWS Bedrock AgentCore Memory. The agent can:
- Answer queries using custom tools (destination lookup, experience info, web search)
- Fetch behavioral attributes from Signals to understand what the user is doing right now
- Retrieve and store customer context across conversations using AgentCore Memory
- Deliver personalized responses that reflect both browsing behavior and historical preferences
This pattern - tools plus behavioral context plus persistent memory - is useful in any customer-facing agent, from support bots to shopping assistants to content recommenders.
If you connected the SEA Explorer demo app, you also have a working frontend that calls Bedrock with Signals and Memory integration, demonstrating the full end-to-end flow.
Clean up
Delete the AgentCore Memory resources created during this accelerator to avoid ongoing costs.
python
from bedrock_agentcore.memory import MemoryClient
from boto3.session import Session
boto_session = Session()
region = boto_session.region_name
memory_client = MemoryClient(region_name=region)
memory_client.gmcp_client.delete_memory(memoryId=memory_id)
print(f"Deleted memory resource: {memory_id}")
If you published a Signals service and attribute group, run the optional cleanup cell at the end of the notebook to remove them:
python
sp_signals.unpublish([travel_service, session_attributes_group])
sp_signals.delete([travel_service, session_attributes_group])
Next steps
- Deploy to production: Use AgentCore Runtime to deploy your agent as a scalable, serverless endpoint with built-in observability. Use AgentCore Gateway to expose your tools as shared MCP endpoints with JWT authentication.
- Add more Signals attributes: Define additional behavioral attributes based on your application's event data - for example, purchase history, content engagement patterns, or feature usage.
- Explore accelerators: The Build a personalized travel agent with Signals accelerator covers a simpler implementation using OpenAI and Signals without persistent memory.
- Learn more about Signals: Read the Signals documentation for advanced attribute definitions and CDI integration.
- Learn more about AgentCore: Explore the AgentCore developer guide and code samples.