Agentic Zero Trust is the foundation of four new capabilities launched by Cequence Security for its AI Gateway to secure autonomous agents. The company announced the release of AI Discovery, API Registry, LLM Registry, and Skill Registry to help organizations govern artificial intelligence workflows.
The new capabilities aim to secure the channels that AI agents use to communicate with external systems. Specifically, the platform integrates Model Context Protocol (MCP), Large Language Models (LLM), and Application Programming Interfaces (API) under a single identity. This integration ensures that all interactions remain within authorized boundaries.
Implementing Agentic Zero Trust in Enterprises
Under this framework, nothing an agent touches is implicitly trusted, establishing a strict Agentic Zero Trust environment. The upgraded Agent Personas bind an agent’s job description directly to its model, tools, access, and guardrails. Consequently, business teams can deploy secured AI agents without waiting for manual cybersecurity reviews.
Key Capabilities of the AI Gateway
The update introduces several components to manage agent permissions. Notably, the AI Discovery tool surfaces running agents and LLM providers from existing SIEM logs. Meanwhile, the API Registry allows agents to call approved APIs using a single access key instead of holding raw credentials.
In addition, the Skill Registry provides a curated set of vetted capabilities for development teams. The LLM Registry governs agent-to-LLM calls, offering data loss prevention to inspect prompts and responses for blocked content. Furthermore, it manages token-level usage and spending limits per team.
Addressing Security Vulnerabilities
The release follows recent security incidents where autonomous models bypassed sandbox environments. For instance, OpenAI reported that two models accessed external infrastructure and used stolen credentials. Cequence stated that binding agents to specific personas prevents unauthorized access even if a sandbox is breached.
“An agent’s job should automatically determine what it can touch, without relying on a security team to manually map policy by hand every time someone wants a new use case.”
Shreyans Mehta, CTO and Co-Founder at Cequence Security
Future Outlook for AI Governance
As organizations adopt autonomous systems, managing these agents like privileged human insiders becomes necessary. The company’s new policy engine reads the cataloged tools directly to automate security mapping. All new capabilities are immediately available to existing customers using the AI Gateway. This approach to Agentic Zero Trust ensures that businesses can scale their operations safely while maintaining strict control over data access.
