A Seattle startup called Certiv emerged from stealth Monday with $4.2 million in funding to build security software that monitors and controls AI agents operating on employee computers.
Certiv is led by CEO Jason Needham, a Seattle tech veteran who co-founded Union Bay Networks, acquired by Apple in 2014, and CloudCoreo, a cloud management platform acquired by VMware in 2018. He also spent more than 12 years at F5 Networks.
The company’s technology addresses the security risks created by AI agents that can now autonomously write code, read files and access sensitive systems using an employee’s credentials.
Its software sits on employee machines (including Windows, Mac and Linux) and acts as a gatekeeper between AI agents and the systems they interact with, evaluating each action against company policies and blocking those that violate them.
Needham’s co-founders are CTO Paul Allen, who co-founded and served as CTO at CloudCoreo and was most recently a distinguished engineer at Broadcom; and Chief AI Officer Daniel Morris, who spent seven years at Microsoft working on AI and developer tools.
Certiv was founded in June 2025 and has nine employees. The company says the funding will support engineering growth and early enterprise deployments. It already has multiple pilot deployments with enterprise customers.
Investors in the pre-seed round included Aviso Ventures, the Seattle-based fund led by Signal Sciences founder Andrew Peterson, along with Founders Co-op, Fortson and others.
The broader market for governing and securing AI agents is active and fast-moving. Mandiant founder Kevin Mandia raised $190 million last week for an AI agent security startup. Israeli startup Onyx Security raised $35 million for a platform to secure AI agents in enterprises. OpenAI acquired cybersecurity startup Promptfoo to strengthen its agent safety tools.
In that landscape, Certiv is staking out a specific niche: intercepting agent behavior on the endpoint, the employee’s computer or device, before it reaches production systems, giving it full visibility into what agents are doing on each machine.
Needham said that lets the company do things that cloud-based or network-level tools cannot: detect unauthorized AI agents running on employee computers, track the full chain of reasoning behind an agent’s actions, and enforce policies based on what an agent is trying to accomplish rather than just what it’s doing at a given moment.
The company calls this “runtime assurance” for AI agents.
“Our fundamental belief is that you cannot control these new workers if you don’t live on the compute where agents actually run,” Needham said.
The company works remotely but spends time regularly at the Foundations startup community on Capitol Hill in Seattle.
