Onyx Platform is an AI control plane designed for enterprise environments to manage, secure, and govern AI agents, models, and applications. It provides centralized visibility and control over AI usage across an organization's infrastructure. The platform is organized around five core capability areas: AI Observability: - Real-time monitoring of AI agent activity across LLMs and frameworks - Full session replay and audit trail for compliance and forensics - Detection of shadow AI usage and unapproved model deployments - Unified view across multi-cloud and multi-agent environments - Anomaly detection and behavioral baselining of AI systems AI Security: - AI Security Posture Management (AI SPM) for configuration hardening - Supply chain risk identification across agents, MCP servers, and models - Automated red teaming to identify vulnerabilities in agents and models - Real-time protection against prompt injection, jailbreaks, and data exfiltration - SIEM/SOAR integration for AI-specific threat intelligence AI Governance: - Policy templates aligned to MITRE, NIST, OWASP, and regulatory frameworks including the EU AI Act - Natural language policy creation for enforceable rules - MCP server access control and tool sanctioning AI Orchestration: - Managed AI gateway for centralizing AI traffic - Intelligent LLM routing with failover, load balancing, and cost optimization - Inline MCP traffic proxy with logging and guardrail enforcement AI ROI: - Executive dashboards tracking AI adoption and return on investment - Department-level productivity metrics, time savings, and cost-benefit analysis The platform includes a supervisory component called the Guardian Agent, which continuously identifies risks and remediates issues across the platform. Deployment options include cloud, hybrid, and self-hosted configurations, with support for AWS VPC, Bedrock Gateway, and custom proxy setups.