Ultra

Ultra provides a runtime security platform for AI agents. The product sits between autonomous agents and the tools or connectors they use, and evaluates each agent action against policy before it runs. The platform is organized around several functions: - Discover: finds agents and connectors in an environment, including those built into tools such as Claude and Cursor, and applies policy automatically when new ones appear. - Observe: intercepts agent actions along with the associated identity, device, connector, and tool, and keeps an audit record of events and the rules that allowed or blocked them. - Govern: lets administrators allow or deny what each agent can do, from entire connectors down to individual tools. - Secure: applies pre-built guardrails against tool poisoning, prompt injection, credential exposure, rug pulls, destructive actions, and PII exposure, with support for custom guardrails. - Detect: raises alerts on anomalous activity such as unusual access, data movement, or tool use, without blocking the action. Ultra can be deployed on individual devices, on employee devices through MDM, and in environments where internal company agents run, with central policy applied across them. Listed customers include Parker, The Martec Group, Vantage, and Michigan State University. The company lists ISO 27001:2022, SOC 2 Type 2, ISO/IEC 42001:2023, and NIST AI RMF alignment on its trust center, and names Firedrop, Entrée Capital, Ludlow Ventures, Hidden Capital, Detroit Venture Partners, and FDVC as investors. Pricing is stated as free to start and pay per action, with no per-seat or per-host fees.