As enterprises move AI agents from experimentation into production environments, security concerns are shifting from model behavior to governance, identity, and authorization. To address those challenges, Tetrate and Ory have announced a strategic partnership focused on securing AI agents at runtime.

The collaboration combines Ory’s identity and authorization platform with Tetrate’s Agent Router Enterprise, creating a framework that applies security policies not only to which tools AI agents can access, but also to how those tools are used. The companies argue that this level of control will become increasingly important as organizations deploy AI agents that can interact with business systems, customer data, and operational workflows.

The Next Security Challenge for AI Agents

Many AI agent frameworks today focus on granting or restricting access to tools. However, security teams are increasingly discovering that tool access alone does not provide sufficient protection. An AI agent may be authorized to use a tool, but the specific actions it takes—or the parameters it submits—can still create risk.

The joint solution aims to address this gap by introducing runtime enforcement at the point where agents interact with enterprise services. Rather than relying solely on static permissions, requests are evaluated in real time against authorization policies.

For example, an AI agent may be allowed to process a customer refund, but a larger-than-normal refund request could trigger additional approval requirements. Similar controls can be applied to financial transactions, healthcare data access, infrastructure changes, or HR-related actions.

“The challenge with AI agents isn’t just controlling which tools they can access—it’s controlling how they use those tools,” said David Wang. “Tetrate Agent Router Enterprise enforces fine-grained authorization on MCP tool invocations down to the parameter level, based on policies defined in Ory, and does so through a globally distributed Envoy-based gateway layer. That gives enterprises the precision, scale and control that production deployments demand.”

Combining Identity and Runtime Enforcement

The partnership brings together two layers of security that are often managed separately.

Ory provides identity, authentication, and authorization services, treating AI agents as first-class identities with assigned permissions and access policies. Tetrate provides runtime enforcement, evaluating live requests as agents interact with models, tools, APIs, and enterprise applications.

When a request exceeds predefined risk thresholds, the system can pause execution and trigger additional authentication or approval workflows. Temporary elevated permissions can then be granted and audited before the action proceeds.

“AI agents must be treated as first-class identities with explicit authentication, authorization and governance,” said Jeff Kukowski. “Together with Tetrate, Ory is helping enterprises secure AI agent deployments end to end, from identity and access decisions to runtime enforcement and policy control.”

The approach aligns with a growing trend in enterprise AI security, where organizations are applying principles such as least-privilege access, zero trust, and continuous verification to autonomous systems.

Built on Open Source Infrastructure

The solution is built on Envoy AI Gateway, an open source project designed to manage and govern AI traffic. Tetrate, a major contributor to the broader Envoy ecosystem, uses the gateway as the traffic enforcement layer for distributed AI environments.

The partnership also reflects an existing customer relationship. Ory previously adopted Tetrate Enterprise Gateway for Envoy to support its own identity platform infrastructure, a deployment that helped streamline operations and improve observability. That experience ultimately evolved into a broader strategic collaboration around AI security.

The combined platform is designed for organizations operating across multiple cloud providers, geographies, and infrastructure environments, where centralized policy enforcement can be difficult to achieve.

What Comes Next

As enterprises increasingly rely on AI agents to automate business processes, security models are evolving beyond simple access controls. Organizations need ways to verify not only who an agent is, but also what it is attempting to do and whether specific actions comply with organizational policy.

The Tetrate-Ory partnership reflects a broader shift in the AI ecosystem toward runtime governance, where identity, authorization, observability, and policy enforcement work together to reduce risk. As AI agents gain access to more sensitive systems and workflows, these controls may become as critical as the models powering the agents themselves.