AIR Raises $50M to Monitor and Secure AI Agent Tools and Components

As organizations grant AI agents increasing access to their internal systems, a new software supply chain is emerging, consisting of the skills, plugins, MCP servers, and add-ons that enable these agents to interact with the internet. Recognizing the need for oversight in this emerging ecosystem, AI security startup AIR has emerged from stealth with $50 million in funding to build a monitoring solution.

Founded by Yair Saban (CEO) and Niv Hoffman (CTO), both veterans of Israel’s Unit 8200 intelligence corps, AIR provides a platform designed to discover agents within a company's network. The platform continuously evaluates the skills, tools, and components these agents utilize and blocks interactions with unauthorized software or external sources. Additionally, AIR operates a marketplace featuring vetted add-ons and skills for AI agents. The funding was raised across two seed rounds totaling $50 million, with the first securing $10 million led by Sequoia and the second bringing in $40 million led by Greenoaks. Participants in the rounds included Swish, Netz, Zach Frankel, Yinon Costica, Ofir Erlich, Anne Neuberger, Omer Adam, Varun Anand, and various angel investors.

Saban argues that the widespread use of AI agents is beginning to resemble an operating system, yet the tools they employ lack the rigorous oversight applied to drivers. He draws a parallel to the evolution of system drivers, stating, "In the early 2000s, whenever you installed a driver, the driver didn’t need to be signed. Today, every time you install a driver, you see a signature saying who signed it... You don’t have that with skills or plugins or MCPs, and it’s a shame, because it’s the same mechanism, it’s the same lesson, but we haven’t learned it."

The primary risk identified by the startup is that as agents become more autonomous, specifically when querying databases and connecting to the web, attackers may poison the content the agents consume rather than directly attacking the system.

AIR addresses this threat through a multi-layered approach. Their visibility product identifies active agents and unauthorized employee tool usage. An enforcement layer intercepts and analyzes agent actions, such as loading a skill or fetching web content, while also checking tools against a maintained whitelist.

To maintain this whitelist, AIR continuously evaluates public skills and add-ons for changes or malicious behavior, as previously approved components can become risky if their dependencies change or developer accounts are compromised. Saban noted that the platform currently filters out approximately 27% of the add-ons and skills found online.

The company reports over 20 customers, with roughly a quarter belonging to large enterprises. Demand is strongest in heavily regulated sectors, particularly financial services and pharmaceuticals.

AIR faces competition from several established players in the space, including Noma Security, which offers discovery and monitoring; Zenity, which provides governance tools; Astrix Security, focusing on identity for agents and MCP servers; and Operant AI, which offers agent protections and an MCP gateway.

Regarding the competitive landscape, Saban believes AIR’s unique advantage lies in its ability to continuously vet the growing ecosystem of skills and add-ons. He stated, "Continuously vetting skills and plugin websites... this is a hard mission to do. Gaining visibility over the endpoint, that is easy. Everybody’s going to do it. It’s hard to create a moat around that." Sequoia Partner Bogomil Balkansky echoed this sentiment, characterizing the challenge as "a continuous re-verification problem" rather than just a scanning issue, noting that "Inspecting every skill, plugin, MCP server and sub-agent an enterprise’s agents touch... is an infrastructure problem."

With around 40 employees, AIR plans to allocate the new capital primarily toward hiring researchers and expanding its go-to-market operations in the United States and Europe.