Using the Step AI Agent on your Step instance
The AI Agent is itself an Automation Package that can be deployed to your Step instance and executed there.
Installation
You need:
- A Step instance running in a SaaS cluster or on-premises.
- A connected Step Node.js agent with Node.js 24 or newer, available throughout the run.
Download the AI Agent package (.jar) from Step Enterprise downloads; request access if you do not have credentials yet. Deploy the JAR to your Step instance; see Upload Automation Package.
Alternatively, select Maven coordinates in the upload dialog and paste this dependency for AI Agent 31.0.1:
<dependency>
<groupId>ch.exense.step-enterprise</groupId>
<artifactId>step-ai-agents</artifactId>
<version>31.0.1</version>
</dependency>
Step downloads the package directly from the repository. This requires Maven repository access to be configured on your Step instance.
Use an Anthropic API key scoped to an Anthropic workspace. On your Step instance, store it as a Step Parameter named ANTHROPIC_API_KEY rather than an environment variable. Set its scope to Global so it reaches every keyword of the run, and mark it Protected so the value is obfuscated in the UI, reports and exports.
Quick start
Deploying the AI Agent puts its plan, AI Agent - Automate, in the instance’s Plans list. You run it from there, like any other plan:
- Open Plans and find
AI Agent - Automate. - Click the play button to open the execution panel.
- Set
instructionsandapDirectory(a writable output folder on the Step Node.js agent where the generated package is saved). See Execution parameters for details. - Run it. The AI Agent’s workflow becomes a Step execution, with one node per phase.
The generated Automation Package is attached to the execution as automation-package.zip.
See Troubleshooting for common execution errors, and Data and privacy for model data handling.