Michigan-based Ultra Security has emerged from stealth with $10 million in seed financing to develop a security platform for autonomous AI agents.
Investors include Ludlow Ventures, Hidden Capital, Entrée Capital, Detroit Venture Partners, Firedrop and FDVC, alongside angel investors from the enterprise software and cybersecurity industries.
Founded in 2025 by Chase Lee, Ultra is targeting a new category of enterprise security infrastructure: controlling what AI agents can access and execute while they operate.
Lee previously founded Trustpage, a security and compliance automation startup acquired by Vanta in 2023. He subsequently served as vice president of new product development at Vanta.
Ultra's founding team also includes engineers and security professionals with experience at companies including Vanta and Secureframe.
Traditional enterprise security tools were largely designed around human users, applications and networks. Autonomous AI agents introduce a different security problem.
Agents can access files, call APIs, retrieve credentials, modify code and execute workflows across multiple systems without requiring human approval for every action.
As businesses increasingly give agents access to operational infrastructure, security teams need to understand not only what data an agent can see, but also what it is doing with that information.
Ultra sits between AI agents and the tools they access through Model Context Protocol connections and other integrations.
The platform intercepts interactions and applies policies before actions are completed.
Its capabilities include preventing unauthorized access to credentials, blocking sensitive information transfers, controlling available tools and recording agent activity for auditing.
Ultra also provides an observability layer that allows security teams to inspect agent actions and investigate potentially unusual behavior.
The product is designed to work across different AI models and agent frameworks, reducing dependence on a specific model provider.
One of its distinguishing features is a local-first architecture. The company says customers can keep agent activity and audit information on their own infrastructure by default, with remote-hosted deployments also available.
Ultra reports achieving SOC 2 Type II, ISO 27001 and ISO 42001 certifications before its public launch. It also says it has completed a FedRAMP 20x Class B audit and is pursuing FedRAMP certification, which should not be confused with already having federal authorization.
The company offers a usage-based pricing model rather than charging exclusively by user seat.
Its published pricing includes a free tier covering up to 5,000 actions per month and a pay-as-you-go tier priced at $1 per 1,000 secured actions, alongside custom enterprise agreements.
That approach links security spending to the number of agent actions being monitored and controlled.
The opportunity is emerging as enterprises transition from AI assistants that primarily generate information to autonomous systems that can make changes across business applications.
However, agent security is quickly becoming a competitive market. Established cybersecurity vendors and specialized startups are developing products addressing agent identity, access management, monitoring and runtime protection.
Ultra's challenge is to become a trusted control layer without introducing latency, implementation complexity or new vulnerabilities.
The company has disclosed early work with customers and design partners but has not provided revenue or customer-count figures.
Its next test will be showing that real-time controls can prevent consequential security incidents while allowing AI agents to remain useful in production environments.