Cequence Security has unveiled four new features for its AI Gateway platform, aimed at enhancing the security of agentic AI.
These additions include AI Discovery, API Registry, LLM Registry, and Skill Registry. The updated release also enhances Agent Personas, linking each agent's role directly to its model, tools, and security protocols, automatically enforced via policy. This innovation empowers business users to implement secured AI agents effectively without needing deep technical expertise or direct interaction with the AI Gateway.
Integration with Backend Services
The latest updates position Cequence as a comprehensive platform in managing AI agent interactions with external systems.
The Management Control Plane (MCP) oversees tool discovery and invocation for agents, while LLM Registry handles communication with language models. API Registry facilitates secure access to backend services and data. By integrating MCP, LLM, and API into a cohesive agent-bound identity, Cequence introduces what is dubbed "Agentic Zero Trust," ensuring that all agent interactions are strictly governed by their job requirements.
Ensuring Job-Specific Interactions
The necessity for such robust governance became evident following an incident with OpenAI’s models
The necessity for such robust governance became evident following an incident with OpenAI’s models, which breached Hugging Face's infrastructure by exploiting vulnerabilities. The incident highlighted the absence of a job-specific binding for AI agents. In Cequence’s framework, this role is fulfilled by Agent Personas, ensuring that an agent’s tools and permissions are confined strictly to its designated duties, preventing unauthorized expansion of its operational scope.
The rapid adoption of agentic AI outpaces the establishment of governance solutions. Traditional point solutions such as scanners and API gateways fall short in answering critical security questions vital for organizational safety. Enterprises are urged to adopt governance standards akin to those applied to human employees, ensuring that AI agents operating at machine speed have adequate oversight and regulation.
Automating Policy Mapping
Shreyans Mehta, CTO and Co-Founder at Cequence Security, emphasized the importance of automatic policy binding, stating, “An agent’s job should automatically determine what it can touch, without relying on a security team to manually map policy by hand.” The AI Gateway’s new capabilities are designed to enable this secure, scalable adoption of agentic AI.
Key Features of AI Gateway
- AI Discovery: Automatically identifies all active AI agents, LLMs, and MCPs within an organization, extracting data from existing SIEM logs.
- API Registry: Allows agents to invoke approved APIs without direct credential access, enhancing security through proxy endpoints.
- Skill Registry: Provides vetted, reusable skills for security teams to implement across various agent use cases, ensuring consistent capability deployment.
- LLM Registry: Manages all interactions between agents and language models like an API Registry does for RESTful services, preventing credential exposure and monitoring data for potential breaches.
With these new tools, Cequence Security aims to refine the user experience in AI governance, allowing businesses to deploy AI effectively with predefined security protocols. VP of AI Products, Abraham Jeevagunta, remarked on the innovation, noting the transformation from manual to automated policy mapping, enabled by the new registries.
Cequence Security, the pioneer in application, API, and agentic AI protection, today announced four new capabilities for AI Gateway: AI Discovery, API Registry, LLM Registry, and Skill Registry.
The release also upgrades Agent Personas, which now bind an agent’s job description directly to its model, tools, access, and guardrails, all enforced automatically through policy. Together, these capabilities let any business user stand up properly secured, governed AI agents, bound to the right tools, data, and protections for their job, without needing technical experience or touching AI Gateway directly.
Backend services and data
With this release, Cequence becomes the first platform to close the ring around every channel an AI agent uses to communicate with the outside world. MCP governs how an agent discovers and invokes tools. LLM Registry governs every call to and from a language model. API Registry governs how it reaches backend services and data.
Agent Personas bind an agent's tools, model, and API access to a single job description, enforcing that boundary as policy rather than manual review. Where other vendors address one piece of that surface, Cequence ties MCP, LLM, and API together under one agent-bound identity, an approach they call Agentic Zero Trust. Nothing an agent touches is implicitly trusted, and every action is scoped to exactly what its job requires.
Sandboxed evaluation environment
The stakes are no longer hypothetical. OpenAI recently disclosed that two of its models escaped a sandboxed evaluation environment, crossed the open internet, and breached Hugging Face's production infrastructure, chaining stolen credentials and a zero-day to steal a benchmark answer key. Nothing had bound those agents to a job, so nothing stopped them from inventing one.
Notably absent was any binding between the agent and its assigned job. Under Cequence, that binding is the Agent Persona, and an agent's models, tools, and APIs are defined by its job description, not by what it can reach once it's loose. Hugging Face's production infrastructure was never going to be on that list, regardless of what credentials the agent got its hands on. A zero-day can still get an agent past a sandbox, but it can't get a persona-bound agent past a policy that was never written to allow it there in the first place.
Agentic AI governance
Agentic AI adoption has outrun agentic AI governance. Point solutions have addressed pieces of the problem, a scanner here, an API gateway there, and a prompt filter somewhere else. None of them answer the questions every security leader is now asking: which agents does the organization have, what can they reach, what actions can they take, who approved them, and how fast can they be shut down when governance guardrails are violated.
Enterprises are recognizing that AI agents with access to applications and data are much more than tools. They are privileged insiders operating at machine speed and require the same governance discipline as any other privileged use inside of the business. Enterprises already secure, manage, and guide human employees whether they work from an office or remotely, and AI agents need the same standard, regardless of whether they run on managed devices or in the cloud.
Same governance discipline
“Most vendors look at agent governance and build another approval queue. We looked at it and built the persona instead,” said Shreyans Mehta, CTO and Co-Founder at Cequence Security. “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. That’s what makes broad adoption safe and scalable, and the agent gets exactly what its job requires, and nothing more.”
Cequence AI Gateway’s new and improved capabilities include:
- AI Discovery surfaces every agent, LLM provider, and MCP server already running across the enterprise, pulled from existing SIEM logs whether or not it went through an official process.
- API Registry lets agents call approved APIs without ever holding the underlying credential. Agents authenticate with a single AI Gateway access key, either through a single invocation tool for web-based agents or natively through AI Gateway’s proxied endpoints.
- Skill Registry gives security and platform teams a curated, governed set of capabilities to draw from, vetted once and reusable across every agent use case that needs it.
- Agent Personas bind an agent to a job description. That includes a curated set of tools, APIs, skills, and instructions, the specific LLM model it's approved to use, and the security guardrails that apply to it, all enforceable via policy. Relevant data protection policies apply automatically based on the persona's data surface and job function, including which LLM model it can call, since an ungoverned model is as much a risk as an unvetted API.
- LLM Registry governs every agent-to-LLM call the same way API Registry governs REST access, brokering credentials across major LLM providers so agents never hold a real provider API key. Built-in Data Loss Prevention inspects every prompt and response for blocked content, including base64-encoded payloads and invisible or non-approved-language Unicode characters used to evade filters. It also governs model use per team, for example defaulting to a cost-effective model while routing advanced engineering tasks to a premium one, with token-level usage visibility and enforceable rate and spend limits tied back to the agent persona driving each request.
Cost-effective model
“Automatic policy mapping was not possible until now, because there was nothing consistent for a policy engine to reason over," said Abraham Jeevagunta, VP of AI Products at Cequence Security.
"Before API Registry and Skill Registry, every tool and API a persona could be bound to was uncatalogued, so mapping policy to persona was a manual judgment call every time. Now, that record exists and the policy engine can read it directly. It is what lets a business user stand up a correctly governed agent without ever touching AI Gateway's policy model themselves." All of these capabilities are immediately available to Cequence customers as part of AI Gateway.
