Onyx Security
Series B
Enterprise secure AI control plane for discovering, governing, and securing AI agents.
Founded
2024Enterprise secure AI control plane for discovering, governing, and securing AI agents.
The Secure AI Control Plane discovers sanctioned and shadow AI agents, monitors prompts, responses, and actions, and enforces security and governance policies in real time. Core modules include AI observability, security and prompt-injection defense, agent identity and access governance, MCP security, orchestration, and AI ROI tracking.
A supervisory Guardian Agent uses proprietary models to identify risks and remediate issues at scale. The product is sold to enterprises that need end-to-end control over third-party copilots and internal agent programs across regulated industries.
Enterprise demand for AI agent control is rising as models gain autonomy and employees grant agents access to critical systems. Security leaders face a new risk class where agents are empowered but not accountable, and traditional chat DLP or static guardrails do not scale to thousands of agents.
Investors and customers are treating secure AI control planes as infrastructure for the agentic era, alongside broader AI security categories such as AI-SPM and runtime defense. Capital and product activity in 2026 point to continued expansion of agent governance spend across banking, energy, healthcare, and technology.
Onyx combines discovery, runtime inspection, and policy enforcement in one control plane that spans SaaS, cloud, endpoints, and code rather than treating agent security as a single-channel add-on. Proprietary models and the Guardian Agent supervisory layer are designed to understand agent reasoning and intervene on risky actions without forcing enterprises to freeze AI adoption.
The platform also layers governance, orchestration, and ROI measurement so security, infrastructure, and AI teams share the same system of record. Deep integrations with major model and agent providers, including Anthropic, support coverage for both third-party and home-grown agents.