AUI

AUI develops Apollo 1, a neuro symbolic foundation for enterprise agents. It combines the language understanding of neural networks with symbolic logic in a single runtime, so a model interprets the request while code enforces policy, performs calculations and decides which actions are permitted. An agent is treated as a versioned program made of typed files that declare facts, capabilities, policies, calculations and connections. Teams describe the work and the rules in natural language in a no code Playground, or author the same program through a developer CLI. They then run it against test scenarios, read an execution trace that shows the facts, rule or service result behind each outcome, revise, and publish a version, with the ability to roll back to an earlier one. The vendor states that no domain retraining or mapping of every workflow is required, and that every decision is traceable. Agents connect to APIs, databases, documents and tool servers through a REST API or a TypeScript and JavaScript SDK. The product is aimed at enterprises that need agents to follow policy deterministically, for example in approvals, reimbursements and other regulated workflows, and want an audit trail for each decision. No prices are published. Access is by request directly from AUI, or through Apollo on the Google Cloud Marketplace.