MagicMirror is an AI observability and protection platform that runs Small Language Models (SLMs) locally in the browser or on local environments to secure and monitor generative AI usage. It provides real-time visibility into AI interactions, detecting shadow AI and enforcing dynamic policies without routing raw prompts or sensitive data to the cloud. The platform tracks AI usage metrics, assesses session-level risks, and offers in-context prompt suggestions to scale effective prompting across an organization. By anonymizing sensitive data on-device before it reaches public Large Language Models (LLMs), MagicMirror enables secure AI adoption while maintaining strict data privacy and user trust.
