Build multi-agent teams that remember every customer with Amazon Bedrock AgentCore

Description and use

Build a multi-agent team that remembers every customer with Amazon Bedrock AgentCore and n8n. One memory harness serves four specialists, each handling different query types while preserving conversation context.

Problem

Customers need a support bot that can handle multiple types of queries (configuration, calculation, architecture, general) while preserving conversation context across agents without building separate vector stores or deploying multiple agents.

Outcome

The workflow demonstrates how to build a multi-agent support team with shared memory, reducing duplication and operational cost. It handles different query types, preserves context across agents, and returns answers in Slack with specialist tags.

Workflow

• Connect AWS and Slack credentials in n8n
• Set Customer Context node to create Actor ID and Session ID
• Triage Agent classifies incoming question and returns category JSON
• Read Triage Decision parses the category
• Route To Specialist sends question to the matching specialist agent
• Each specialist agent runs with tools granted per call (Code Interpreter, AWS skills catalog, shell)
• Format Reply unwraps the answer
• Post Answer To Slack posts the answer tagged with the specialist

Evidence

AgentCore’s Managed memory is scoped by actor and session, so an Actor ID that identifies the customer rather than any single agent gives every specialist the same history to read and write

Source

The workflow creates one harness for all four agents rather than one each

Source