n8n Workflow: Per-Customer Memory with Amazon Bedrock AgentCore

Turn n8n chat into a support assistant that remembers each customer—across sessions

This n8n workflow uses an Amazon Bedrock AgentCore support agent with managed long-term memory, so your assistant can maintain per-customer context between separate chat sessions and return a clean, safe plain-text response to the n8n chat interface.

What this workflow does

• Receives each incoming message from the n8n chat trigger as a separate execution.
• Sets the agent name and resolves a customer identifier (by default, it uses the chat session ID—ideal for testing).
• Passes the user’s message into Amazon Bedrock AgentCore while enabling managed memory.
• Scopes memory by customer and session using:
  • Long-term memory: keyed to actor ID (customer)
  • Continuity: keyed to session ID
• Parses and normalizes the AgentCore response (including optionally fenced JSON) and returns a plain-text output suitable for chat.

Use cases

• Customer support with continuity: recall prior requests and preferences per user even after they leave and return.
• Multi-tenant SaaS assistance: map customerId to your real user ID so each account has its own memory.
• Operational helpdesk bots: maintain context over time without manually storing conversation history.

Technical details

• Nodes / components: set, code, sticky note, n8nn8n-nodes-langchainchat chat trigger, and awsn8n-nodes-agentcore AgentCore harness.
• Amazon Bedrock: requires an AgentCore API credential (AWS access key, secret, region, and execution role ARN).
• Setup step: update the customerId mapping to your real user identifier (or keep the default chat session ID for testing).
• Optional configuration: customize the AgentCore system prompt and agent name to match your support policy and tone.