Company Interview / JFrog teams up with Nvidia to manage AI agents

Key Points:

Rao highlights rising risk from “shadow AI” inside enterprises AI supply chains add models, tools, MCP servers and skills to traditional software risk JFrog’s agent skills registry integrates first with Nvidia to act as a governance control plane Rao expects AI security and governance to be fully embedded into development workflows within 2–3 years

Sunny Rao from JFrog sets out a roadmap for enterprises grappling with the rapid adoption of autonomous AI agents. Rao states that organisations are racing to harness AI for productivity and cost reduction, yet many lack visibility over the AI tools and agents already operating inside their environments. In his view, this “shadow AI” creates operational, regulatory and compliance risk, particularly when boards demand proof of controls, audit trails and incident response.

Rao argues that AI supply chains resemble traditional software supply chains but add new layers of complexity, including rapidly evolving models, tools, MCP servers and task-specific “skills” such as AI wealth managers. He positions JFrog’s new agent skills registry, launched with Nvidia as its first integration partner, as a “system of record” and control plane designed to provide end‑to‑end visibility, metadata management and policy enforcement over AI models and agents.

Looking ahead two to three years, Rao expects AI security and governance to be embedded into every development workstream, with verification and compliance checks running automatically at machine scale. He views partnerships with leaders such as Nvidia and broader marketplace availability as critical to enabling enterprises to deploy AI at scale while maintaining trust, safety and regulatory readiness.