Onyx Raises $113M to Bring AI Agent Governance to SaaS Stacks

As every major SaaS vendor — from Salesforce to ServiceNow — rolls out AI copilots and agents, the operational surface area for security teams has exploded. Onyx Security’s platform specifically targets this gap, offering real-time monitoring of AI agent behavior within SaaS workflows. For CIOs and CISOs at SaaS-heavy enterprises, the ability to discover shadow AI, block prompt injections, and ensure every agent action aligns with policy is no longer optional — it’s the difference between safe acceleration and an unmanageable risk crisis. The $113 million infusion means the platform will likely expand its integrations deeper into the SaaS ecosystem, becoming a crucial control plane.

Onyx Security, an Israeli startup, has closed a $113 million Series B funding round, bringing its total capital raised to $153 million and reportedly valuing the company at an estimated $640 million. The round was led by Bessemer Venture Partners, with strong participation from Cyberstarts, TCV, Conviction, FirstMark, Vintage, QuantumLight, and G Squared — a syndicate that blends top-tier security-focused VCs with growth-stage generalists. The funding marks a pivotal moment for the emerging market of AI agent governance, as enterprises increasingly deploy autonomous AI agents that interact with sensitive corporate systems but lack inherent accountability or oversight.

“

Onyx Security, an Israeli startup, has closed a $113 million Series B funding round, bringing its total capital raised to $153 million and reportedly valuing the company at an estimated $640 million.

The core challenge Onyx addresses is the proliferation of “shadow AI” — unsanctioned or unmonitored AI tools creeping into enterprise environments, from SaaS applications to cloud infrastructure. These AI agents, while boosting productivity, can introduce severe risks: they may execute unintended actions, fall victim to prompt injection attacks, or violate regulatory mandates like GDPR or sector-specific compliance rules. Onyx’s platform acts as a control layer, continuously monitoring AI agent decision-making steps across SaaS, cloud, and endpoint environments. By leveraging proprietary models, it can trace an agent’s reasoning in real time, detect anomalous or policy-violating behavior, and intervene automatically — much like a guardrail system for highly autonomous software.

The timing is strategic. Enterprise AI adoption is accelerating, with agentic architectures topping CIO priority lists, yet security frameworks remain nascent. Traditional endpoint detection and response (EDR) or identity tools are ill-equipped to govern a copilot that books travel or a coding agent that pushes repos. Onyx positions itself as the missing infrastructure, effectively redefining the security stack for the AI era. The involvement of Cyberstarts — a firm known for backing foundational cybersecurity categories — signals investor conviction that AI agent control could become as ubiquitous as cloud access security brokers (CASBs) became in the last decade.

From a market perspective, the $113 million raise is one of the largest Series B rounds in the cybersecurity sector this year, and it follows a Series A that closed in 2024. The valuation of $640 million implies strong revenue multiples, likely predicated on early enterprise traction. While Onyx hasn't disclosed customer count or revenue, the willingness of Bessemer and TCV to co-invest suggests significant pipeline visibility. TCV’s presence in particular hints at potential for a near-term IPO path, given the firm’s track record of backing pre-IPO enterprise companies.