Security teams have spent two decades buying tools that find problems. Northstar Security AI launched with $7 million to handle what follows, arguing that context, not detection, is the missing layer.
Existing security and IT tools feed a live Security Context Graph. An unpatched flaw, an unmanaged device and an overprivileged account each look harmless alone. Together they can leave a critical system wide open, because the records sit in products that never talk.
Six tools and one laptop
Northstar’s demo starts with a laptop lacking endpoint protection. Deciding if that machine is dangerous may take six products. Ask in ordinary language and the answer comes back with a twist, since the laptop belongs to a privileged user with a path to payroll.
Built-in agents then push endpoint protection onto the machine while people supervise where it matters. Customers can wire in third-party agents through the Model Context Protocol, Anthropic’s Claude among them. The platform hooks into over 70 security and IT products, including Okta, GitHub and Workday.
A crowded runway
Co-founder and chief executive Amer Deeba argues that teams are awash in signals yet short on the context that makes them actionable, and that decades of tool spending left follow-up manual. That is a fair diagnosis, and among the most contested claims in security right now, since every vendor pitches an agentic remediation layer.
Wing Venture Capital led the seed, with MVP Ventures and Outliers Venture Capital joining and three veterans investing personally, Gerhard Eschelbeck, Renaud Deraison and Justin Somaini. Deeba started the company with chief technology officer Ed Rowe, a former Adobe colleague.
The seed is small against that competition. Northstar bets that model-agnostic wiring into 70 systems buys time, and that buyers tired of agent lock-in will pay for a neutral layer.
