Harness ARN

A Harness ARN is a unique address that points to a specific AI agent. Just as every mailbox has its own street address so a letter sent there reaches exactly that house, knowing this value lets you call back the exact same agent out of many others.

Here, "harness" refers to the managed framework that works behind the scenes so an agent can reason on its own, use tools, and remember previous conversations. Think of it like a climbing harness that keeps an agent moving safely. In automation tools like n8n, there's a single field for this address value: leave it blank and the first run creates a new agent and issues that address; if you already have an issued address, entering it lets you keep using the same existing agent.

In other words, this one field alone determines whether a new agent is created or an existing one is reused. It's also all you need if you want to call up, from another automation tool, an agent that a collaborating team already built.

The article explains that "the node decides how it behaves based on a single Harness ARN field." Leaving the field blank means it will be auto-generated, while filling it in calls up an existing agent. Contrary to a common misunderstanding, this value isn't a model name or configuration detail — it's simply an address that identifies that particular agent.

When adding a Bedrock AgentCore node in n8n, try running it once with the Harness ARN field left blank. After the run finishes, check whether an auto-generated address value appears somewhere, copy it, and enter it into the same field on the next run. Comparing whether the second run picks up the same memory or state as the first will make the role of this field click.