Citrix adds NetScaler MCP Gateway controls for LLM and agentic AI traffic

Citrix announced new NetScaler MCP Gateway functionality to help enterprises govern traffic between AI clients and backend Model Context Protocol servers.

Citrix targets governance for agentic AI traffic

Citrix said in its July 2026 announcement that NetScaler MCP Gateway is intended to securely route, govern, and observe agent traffic to backend MCP servers. The company positioned the feature as part of a broader effort to give enterprises a control point for AI systems that connect to tools, data sources, and services through the Model Context Protocol.

NetScaler documentation describes MCP Gateway as a centralized control layer between MCP clients and backend MCP servers. According to the documentation, the gateway handles connectivity, credentials, and routing logic, reducing the need for each AI client to manage those functions separately.

The announcement focuses on a practical enterprise concern: as AI applications become more capable of taking actions through connected systems, companies need ways to authenticate access, enforce policy, and observe what those systems are doing. Citrix’s stated approach is to place governance and routing functions in the network layer through NetScaler.

Centralized authentication, controls, and routing

APMdigest reported that the NetScaler MCP Gateway capabilities include centralized authentication, granular controls, rate limiting, allow and block lists, and governance for agentic AI traffic. Those functions are designed to help organizations define which AI clients can reach which MCP servers and under what conditions.

Citrix’s own announcement says the gateway can help route and manage agent traffic to backend MCP servers. NetScaler documentation adds that the gateway manages credentials and routing logic, which can be important in environments where multiple internal services, data sources, or tools are exposed to AI applications.

The architecture described by Citrix and NetScaler documentation suggests a model in which enterprises do not rely solely on controls embedded in individual AI applications. Instead, they can use a centralized gateway to apply policies across connections between MCP clients and MCP servers.

AI Gateway adds model routing and usage tracking

Citrix also announced enhancements to NetScaler AI Gateway. According to Citrix, those enhancements include model routing and token-level usage tracking for large language model traffic. MarketScreener also corroborated that Citrix introduced AI Gateway updates covering model routing and token-level usage tracking.

Model routing can help organizations direct requests to different AI models based on policy or operational needs. Token-level usage tracking gives administrators a way to monitor consumption of LLM services, which can matter for cost management, auditing, and capacity planning.

Citrix described these capabilities alongside the MCP Gateway announcement, indicating that the company is addressing both traditional LLM traffic and newer agentic AI patterns. In the sources provided, Citrix frames NetScaler as a governance layer for AI application traffic rather than as a model provider.

Why it matters for enterprise AI deployments

The Model Context Protocol has become a common reference point for connecting AI applications with external tools and data. In enterprise settings, those connections raise security and operational questions: which systems can an AI client access, how are credentials handled, and how can administrators observe traffic patterns?

Citrix’s NetScaler MCP Gateway is aimed at those questions by centralizing connectivity, credentials, routing, and policy enforcement between MCP clients and servers, according to NetScaler documentation and Citrix’s announcement. The related AI Gateway enhancements add controls for LLM traffic, including model routing and token-level tracking.

The significance of the announcement is not that it changes the underlying capabilities of AI models. Rather, Citrix is adding infrastructure controls around how AI applications connect to enterprise systems and consume model services. For organizations experimenting with agentic AI, that governance layer may become a required part of moving from prototypes to managed production deployments.