Summary
- Amazon Bedrock AgentCore harness has reached general availability (GA)
- It integrates directly into the editor via the open-source community node @aws/n8n-nodes-agentcore for n8n
- Persistent memory, code execution, VPC isolation, and more can be configured without writing infrastructure code
AWS announced that Amazon Bedrock AgentCore harness, which allows users to build production-grade AI agents within n8n workflows, is now generally available. While n8n's existing AI Agent node was limited to single model calls, the newly released open-source community node @aws/n8n-nodes-agentcore enables configuration of persistent memory, usable tools (browser, code sandbox), and per-session isolated environments directly within the editor.
What was released
The AgentCore harness provides the "backbone" needed to run agents—orchestration loop execution, tool invocation, context management, cross-session state persistence, failure recovery, and session isolation—as a managed capability. Users simply define the model, tools, skills, and instructions as configuration values, and if configuration alone isn't enough, they can export to Strands code to continue building.
The node's behavior is determined by a single Harness ARN field. Leaving the field empty causes an agent to be auto-created on first run and reused/updated afterward, while entering an existing ARN allows calling an agent built outside of n8n directly. The credential method matches the existing n8n AWS Lambda and Amazon S3 nodes.
What changes
Beyond Amazon Bedrock, the node is compatible with OpenAI, Google Gemini, and LiteLLM-supported providers, and the model provider can be switched turn by turn within the same conversation. AWS noted that the node can be found and installed as a "verified community node" in n8n's node panel, and is released under the MIT license.
Comments