AWS's new AgentCore node brings production AI agents to n8n

AWS has made Amazon Bedrock AgentCore harness generally available, and a new open-source community node, @aws/n8n-nodes-agentcore, integrates it directly into n8n's no-code workflow editor. The node acts as a managed orchestration layer, handling the agent loop, tool calls, context management, state persistence, and error recovery so you don't have to build that infrastructure yourself. It's MIT-licensed and powered by Strands Agents, AWS's open-source agent framework.

With this node, n8n users can deploy agents that maintain memory across sessions, scope memory to individual users, use a code interpreter and web browser as tools, and switch between model providers—including Amazon Bedrock, OpenAI, Google Gemini, and LiteLLM-supported providers—even between turns in the same conversation. The single "Harness ARN" field simplifies setup: leave it blank and the node creates and manages the agent on first run; paste an existing ARN to invoke agents built outside n8n. The node uses the same AWS credential pattern as n8n's Lambda and S3 nodes, and supports running agents privately in your own VPC.