Cyris

Cyris provides an automatically generated audit trail for decisions made by AI agents, capturing inputs, outputs, timestamps, and the originating model without requiring code changes. By installing a two‑line SDK, it intercepts calls to services such as OpenAI, Anthropic, Azure OpenAI, AWS Bedrock, Google Gemini, and others, producing per‑decision logs, real‑time anomaly alerts, and patient‑level workflow traces that can be queried via a URL.

The system is aimed at AI‑native healthcare startups that must demonstrate explainability, data lineage, and compliance to hospital security reviews. It discovers active agents from live traffic, maps handoffs between them, and constructs a workflow graph that satisfies audit questionnaires, while also flagging protected health information before it leaves the process.

Cyris delivers drift metrics, incident histories, and exportable audit data, enabling engineers to monitor decisions, catch problematic outputs before clinicians see them, and respond quickly to compliance requests without additional instrumentation effort.