Northstar Security Exits stealth with $7M to Expose toxic risk combinations

Northstar Security emerged from stealth on Oct. 28, 2026, with $7 million in seed funding to launch an AI-agent platform that maps toxic combinations of identities, assets and vulnerabilities. Wing Venture Capital led the round, with MVP, Outliers and former Google CISO Gerhard Eschelbeck, plus Renaud Deraison and Justin Somaini, also investing. The San Francisco company built its platform on a real-time Security Context Graph that connects assets, identities and applications with vulnerabilities and business systems.

The platform targets what Northstar terms toxic combinations: cases where a vulnerability, an unmanaged device or an overprivileged identity seems harmless alone but together expose critical business systems. The announcement says enterprises have spent billions on tools to find threats, vulnerabilities and misconfigurations, but critical security information stays fragmented across disconnected systems, and tracking down these combinations requires analysts to pull information from multiple tools, assess relationships, judge business impact and arrange fixes, work that runs for hours and buries security teams in manual effort.

The graph updates in real time, showing how assets, identities and applications relate to vulnerabilities and business systems. Analysts can pose questions in plain language to dig into findings, and the platform’s AI agents can kick off remediation workflows, such as deploying endpoint protection, notifying the owner, or restricting risky access, with human oversight where appropriate. The platform provides three capabilities: connected security context, context-driven risk intelligence and agentic remediation.

The platform supports more than 70 enterprise security and IT integrations, including AWS, Google Cloud, Microsoft Defender, Microsoft Entra, Okta, GitHub, Workday, Slack and Zendesk. Customers can allow external AI agents to query the graph through the Model Context Protocol and build their own security workflows with Claude and other AI tools. The platform is available, and organizations can request a demonstration at northstar.security/request-a-demo.

Amer Deeba is CEO and co-founder, and Ed Rowe is CTO and co-founder. Deeba previously co-founded Normalyze, which Proofpoint acquired in 2024, and held senior roles at Qualys. Rowe was CTO at Farmers Business Network and served as a senior vice president leading engineering at AppDynamics. The two founders worked together at Adobe in the late 1990s, and the announcement credits the founding team with decades of experience across security platforms, enterprise software and large-scale data infrastructure.

The funding will support platform development, new hires and go-to-market expansion. Amer Deeba, CEO and co-founder, said in the announcement, “Security teams are drowning in data but starving for context.” Randy Barr, CISO at Cequence Security and a named customer using Northstar’s agentic workflows for remediation, said teams do not need another dashboard that generates more findings; they must connect existing data, understand what matters, and turn context into action. Wing Venture Capital partner Jake Flomenberg said the industry spent billions on better ways to find problems while security teams struggle with uncovered risks, and he added, “We believe it has the potential to fundamentally rewrite the security playbook.”

Pomegra observes that vendors promote graph-based correlation in identity security, cloud security posture and exposure management, which are also adding AI agents. Northstar has not disclosed its valuation, headcount, customer counts or revenue. The outlet added that the $7 million seed is modest next to late-stage cybersecurity rounds, and that the founders’ data-graph background likely explains why Wing led the round. The outlet warned that alert volume is a common complaint among security teams, and that an agent triaging on partial data might multiply noise rather than reduce it; Northstar is betting a unified graph gives agents the cross-tool view human analysts currently piece together by hand.

Pomegra highlighted the tests ahead: winning customers past early adopters and enterprise acceptance of third-party agents querying a central security graph through the protocol. The outlet said opening the graph allows customers to build workflows on Anthropic’s Claude and other AI tools, which turns the graph into a data layer as much as a finished product, and that the harder sell for Northstar is trust in automated remediation rather than graph construction. Pomegra outlined two possible futures: a Series A linked to production deployments using autonomous remediation, or large platform vendors taking over graph correlation and shrinking the independent market, both requiring published evidence of measurable risk-reduction gains.