AIR Raises $50 Million to Secure AI Agent Skills and Add-Ons

As companies give AI agents access to more internal systems, a new security challenge is emerging around the software components those agents rely on, including skills, plugins, MCP servers and other add-ons. AI security startup AIR believes enterprises will need better ways to monitor and verify this growing ecosystem.

The company is coming out of stealth with $50 million raised across two seed funding rounds to build a platform designed to help organisations discover AI agents operating inside their environments, evaluate the tools those agents use and block unsafe interactions.

AIR was founded by CEO Yair Saban and CTO Niv Hoffman, who previously worked in Israel’s Unit 8200 intelligence corps on offensive cybersecurity. The company’s platform is designed to continuously assess AI agent skills, tools and components while providing access to a marketplace of security-reviewed add-ons.

Building security controls for AI agent tools

AIR’s founders argue that AI agent tools are becoming similar to software components installed on traditional systems. Still, they currently lack the same level of verification and oversight that other technologies receive.

The company says the main risk is not only attacks against AI agents themselves, but also attacks against the information and tools they consume. As agents become more autonomous and connect with business systems, attackers could attempt to compromise the skills, plugins or external resources used by those agents.

AIR’s platform is designed to provide visibility into AI agents running across an organisation, identify employees who are using unauthorised AI tools, and monitor agent actions. The company said its enforcement layer can analyse activities such as loading skills or retrieving content from external sources.

The startup also maintains a security review system for AI agent tools and add-ons. AIR said previously approved components can become risky if their dependencies change or if the developer accounts behind them are compromised, making continuous verification necessary.

Growing competition in AI agent security

AIR said its platform currently filters out about 27% of the AI agent skills and add-ons it discovers online. The company claims to have more than 20 customers, with about one-quarter coming from large enterprises, particularly in regulated industries such as financial services and pharmaceuticals.

The company is entering a competitive market with several other startups building security and governance tools for AI agents. Noma Security provides agent discovery, access controls and runtime monitoring, while Zenity focuses on AI security governance and controls.

Astrix Security offers tools to discover and manage AI agents and MCP servers, while Operant AI protects AI agents and MCP infrastructure.

The sector has attracted significant venture investment, with other AI security companies also raising large funding rounds as enterprises look for ways to manage the risks associated with expanding AI adoption.

A focus on continuous AI security verification

AIR says its focus is different from traditional security scanning because AI agent tools can change over time. The company argues that organisations need continuous re-verification of skills, plugins, MCP servers and other components used by their AI systems.

The new funding will support AIR’s hiring efforts, including expanding its research team and growing its go-to-market operations in the United States and Europe. The company currently has around 40 employees.