Pomegra
Ultra Security exited stealth with a $10M seed round to monitor and control autonomous AI agents at runtime, with Ludlow, Entrée and Hidden Capital on board.
Key Takeaways
- Ultra Security emerged from stealth on October 21, 2026 with a $10M seed round.
- Ludlow Ventures, Hidden Capital, Entrée Capital, Detroit Venture Partners, firedrop and FDVC participated.
- The platform sits between AI agents and the MCP servers they call, and it runs locally by default.
Lead
Ultra Security, a Michigan-based startup founded in 2025, came out of stealth on October 21, 2026 with a $10 million seed round. The company sells a platform that watches what autonomous AI agents do and enforces rules on those actions as they happen. Chase Lee, who founded the compliance startup Trustpage before its acquisition by Vanta, is the founder and CEO. The company lists its headquarters as Birmingham, Michigan.
What Does Ultra Security Actually Do?
Ultra puts a checkpoint between AI agents and the MCP servers and connectors those agents depend on. Every tool call, data read and prompt request passes through it, so a security team can see agent activity and block it while it is under way instead of reviewing logs afterward.
The product has four parts. Guardrails enforce policy in real time, including blocking credential access, redacting personal data, protecting health information and limiting data movement between connectors. Governance controls decide which agents, connectors and tools a given user or team can reach. Observability logs and traces agent-to-tool interactions and applies AI-based anomaly detection. A universal connectivity layer lets one setup serve multiple vendors and models.
The software runs locally on user machines by default. That choice matters for buyers wary of routing agent traffic, which may include sensitive data, through another vendor's cloud. Ultra also says its policies can be written in plain language.
Who Backs the Round, and What Is Missing?
Six institutional investors took part: Ludlow Ventures, Hidden Capital, Entrée Capital, Detroit Venture Partners, firedrop and FDVC. Angels from the security and SaaS communities also joined. The announcement names no single lead, and the valuation is undisclosed. The company has not published customer names, headcount or a breakdown of how it will spend the money.
A $10M seed is large for a company founded roughly a year before its public debut. Lee's earlier exit to Vanta and a founding team drawn from Vanta and Secureframe likely explain part of the investor appetite. Both are compliance-automation companies, which fits Ultra's early emphasis on audit credentials.
Why Are AI Agents a Security Problem Now?
Agents now hold real access to corporate systems, and conventional monitoring was not built to watch them. Through MCP connections, an agent can read email, touch code repositories and call internal tools without a person approving each step. Tools that flag problems after the fact offer little help when an agent can move data in seconds.
That gap is the opening for a new product category. Established endpoint, identity and data-loss vendors are adding agent features, but Ultra argues that a platform built only for autonomous agents will outperform an add-on. That claim is untested at scale, and no public customer evidence supports it yet.
How Does Ultra Position Its Compliance Credentials?
Ultra says it holds SOC 2 Type II, ISO 27001 and ISO 42001 certifications, the last of which covers AI management systems. It reports conformance with the AARM Extended specification and alignment with the NIST AI Risk Management Framework. It has also completed a FedRAMP 20x Class B audit and is pursuing full FedRAMP certification.
For a seed-stage company, that list is unusually long. Security buyers often stall on vendor paperwork, so early certifications can shorten procurement. They do not show that the detection works, though, and they say nothing about performance against real attacks on agent workflows.
What Comes Next?
The main test is whether enterprises accept a local-first checkpoint as their control point for agent traffic. If they do, Ultra competes on policy quality and integrations. If large platform vendors bundle similar controls into existing suites, a standalone startup will need clear technical advantages to hold its position.
Funding in agent security is also accelerating, which will bring direct competitors and pressure on pricing. Ultra's next disclosures to watch are named customers, a valuation or follow-on round, and any independent testing of its anomaly detection.
Outlook
Ultra Security enters the market with $10M, a founder who has already sold a company, and a product aimed at a real gap: agents acting on corporate systems with little live oversight. The undisclosed valuation, the absence of a named lead and the lack of customer data leave the commercial case unproven. The next twelve months should show whether runtime control of agents becomes a standalone category or a feature absorbed by larger security vendors.
