AIR raises $50M to vet AI agent plugins

AIR emerged from stealth on September 1 with $50 million in seed funding: Sequoia Capital led an initial $10 million round, and Greenoaks Capital Partners led a further $40 million, with Swish Ventures and Netz Capital also participating. The Israeli company was founded by CEO Yair Saban and CTO Niv Hoffman, both veterans of Israel’s Unit 8200 intelligence corps.

AIR’s product is a real-time firewall built specifically for AI agents rather than for the applications around them. It discovers which agents are already running inside a company, a category most security teams currently have no reliable inventory of, then continuously vets the skills and plugins those agents pull in and blocks anything that looks suspicious. AIR says its own research found thousands of AI add-ons in circulation that rely on untrusted or unverified sources, the specific gap it built the product to close.

The size of the round, six months after founding, reflects investor conviction that agentic AI has opened a security hole existing tools cannot see: software built to scan static code has no equivalent for an agent that decides at runtime which third-party skill or MCP server to call. AIR’s launch lands the same week as Huskeys’ $27 million round for a related but distinct problem, policing the network traffic AI agents generate rather than the plugins they load, and follows MCP’s enterprise-grade managed-auth spec reaching stable, the same connector layer AIR is vetting. Both funding rounds point at the same conclusion for builders: if your team lets agents install skills or connect to MCP servers with no formal vetting step, that is now a funded, named category of attack surface, not a hypothetical one.