AI security startup AIR has emerged from stealth with $50 million raised across two seed rounds. The company is building controls for the growing collection of skills, plugins, Model Context Protocol servers and other components that let workplace agents reach external systems.
AIR says its platform discovers agents running inside a company, including unapproved tools or personal accounts, then intercepts actions such as loading a skill or retrieving internet content. It continuously checks available components and can block those outside an approved list. That repeated review is important because a previously safe plugin can change or begin downloading a compromised package. The company says it rejects about 27% of the add-ons and skills it finds online.
Sequoia led a $10 million round and Greenoaks led the subsequent $40 million financing. AIR reports more than 20 customers, roughly a quarter of them large enterprises, with early demand concentrated in finance and pharmaceuticals. Competitors including Noma Security, Zenity, Astrix Security and Operant AI address similar agent-governance risks, so AIR's central claim is that continuous ecosystem-wide rechecking—not basic agent discovery—will differentiate its service.