Author
Describe the task and policies. Work in the no-code Playground or through the developer CLI.
BUILD
An agent is a versioned program: typed files declaring facts, capabilities, policies, calculations, and connections. Author it, run it against scenarios, read the trace, and revise. The foundation stays the same as the program changes.
Describe the task and policies. Work in the no-code Playground or through the developer CLI.
Test conversations and established cases against the declared behavior.
Read the trace. Find the facts, rule, or service result behind the outcome.
Make a reviewable change. Re-test, publish a version, or republish an earlier one to roll back.
In the Playground, describe changes in natural language and inspect traces. With the CLI, developers and coding agents author and test the same program in their existing tools. Update just the behavior you want to change, without domain retraining or a model release. Explicit policies and traceable changes reduce the work of maintaining the agent as requirements evolve.
APIs, databases, documents, and tool servers supply facts and execute permitted actions. Integrate using the REST API or TypeScript/JavaScript SDK. The same rules can serve different languages; regional policy changes stay explicit.
GET ACCESS
Connect with AUI directly, or reach Apollo through the cloud environment you already use.
BEFORE YOU BUILD
Have a deployment or commercial question?
Talk to the AUI team